1 INFO-VAX	Fri, 28 Jul 2000	Volume 2000 : Issue 419       Contents:% Re: 8400 aka GS140, is it or isn't it  Re: Apache for OpenVMS (VAX?)  Re: BACKUP COMMAND$ Re: CETS2000 - Very Important Notice; Datatrieve (was Re : Interbase 6.0 released as Open Source) ) Re: Exabyte Mammoth-2 tape drives and VMS ) Re: Exabyte Mammoth-2 tape drives and VMS ) Re: Exabyte Mammoth-2 tape drives and VMS  Re: Fortune cookies? GTECH DK hiring ! help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings % Re: help with MACRO compiler warnings ) Re: I/O Size and I/Os per second overview ) Re: I/O Size and I/Os per second overview . Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?. Re: I/Os per second, how do you increase them?) Re: Interbase 6.0 released as Open Source ) Re: Interbase 6.0 released as Open Source ) Re: Interbase 6.0 released as Open Source / Re: is there a database available for VMS hobby / Re: is there a database available for VMS hobby  Re: Memory & CPU Support  Re: Microvax 3100-10 and 20  $99  Re: Microvax 3100-10 and 20  $99  Re: Microvax 3100-10 and 20  $99 Re: Microvax 3100-10 and 20 $99  Re: Microvax 3100-10 and 20 $99  Re: MS IntelliMouse on Alpha?  new release of CRINOID OpenVMS on an Alphaserver 2100" Re: OpenVMS on an Alphaserver 2100" RE: OpenVMS on an Alphaserver 2100" RE: OpenVMS on an Alphaserver 2100" Re: OpenVMS on an Alphaserver 2100" Re: OpenVMS on an Alphaserver 2100" Re: OpenVMS on an Alphaserver 2100) Re: Patwhorks Performance and TCP NODELAY ) Re: Patwhorks Performance and TCP NODELAY ) Re: Patwhorks Performance and TCP NODELAY  performance ratings  RE: performance ratings  Re: performance ratings = Re: Sending a HTML content e-mail from VMS 5.5-2 using MX 4.2 ( Re: Shades of "Nothing sucks like a VAX"( Re: Shades of "Nothing sucks like a VAX" Re: Standalone backups no more?  Re: Standalone backups no more? ! Re: Stirrings in the undergrowth? 
 Re: Such Crap 
 Re: Such Crap  Re: Updating RMS Files Re: Updating RMS Files  URGENT: VMS keyboard on SUN etc!$ Re: URGENT: VMS keyboard on SUN etc! Re: Using CMS and NFS : Re: VMS process manipulate data on SQLServer/NT box : How?< Re: WAY OFF TOPIC: help me with the seed of satan pseudo-O/S
 WTB: RSX Book   F ----------------------------------------------------------------------  # Date: Fri, 28 Jul 2000 11:54:09 GMT / From: "John Nixon" <jorlnixon@worldnet.att.net> . Subject: Re: 8400 aka GS140, is it or isn't itE Message-ID: <Bveg5.1200$ZL5.63416@bgtnsc07-news.ops.worldnet.att.net>   H Thanks Hoff, Curtis, and those who contacted me off line.  I now think I
 understandK the difference, and I can stand in front of our buracuracy and explain what  I wantL with a clear understanding.  When I can do that, I ususally get what I want. When' they see that I am confused, I am dead.   L And thanks for pointing me to the GS80.  It sounds exactly like what I need. ItJ seems to have the performance advantage of the ES40, and more PCI slots to< allow me the types of expansion  I need, and it is scalable.  J What I don't understand, (but am not complaining about) is why the ES40 is onlyL rated as a workgroup class machine and the GS80 is a departmental machine. II guess it has more to do with the number of "things" you can connect to it  rather" than the amount of work it can do.  ? "Hoff Hoffman" <hoffman@xdelta.zko.dec.nospam> wrote in message 0 news:8lq275$icj$1@mailint03.im.hou.compaq.com... > J > In article <28Xf5.18347$o71.1178026@bgtnsc04-news.ops.worldnet.att.net>,1 "John Nixon" <jorlnixon@worldnet.att.net> writes: K > :If I put EV6/525 cpu boards in an Alpha 8400, does that make it a GS140?  > K >   Marketing and services may well have a different perspective on exactly F >   what makes an AlphaServer GS140 series system an AlphaServer GS140 seriesL >   system (and the question certainly has its share of rather philosophicalA >   overtones :-), but strictly from a technical perspective, the  AlphaServer L >   GS140 with the 21264 or later (EV6 or EV67) CPU modules installed is theH >   same hardware platform the AlphaServer 8400 series and its 21264 and# >   21264A (EV5, EV56) CPU modules.  > A >   The central difference across the series of AlphaServer 8200,  AlphaServer L >   8400, AlphaServer GS60, AlphaServer GS60e, and AlphaServer GS140 systemsI >   involves the particular CPU modules used, and differing mounting, and J >   differing I/O and power and power expansion options.  The TLSB (systemI >   bus), memory, I/O and various other core components are all the same.  > J >   There may well be some differing support issues involved, and it wouldJ >   not surprise me to learn that the TL6 CPU modules required a different >   number of license units. > E >   Depending on the particular bottleneck encountered on the current J >   AlphaServer 8400 series (and obviously I know very little of , you mayG >   or may not gain with the AlphaServer GS140 upgrade.   Much like any J >   other upgrade, you will need to upgrade the component that is limitingH >   your current performance in order to see a performance increase.  AsJ >   I am sure you are also already considering, I'd also take a look at anJ >   upgrade to the AlphaServer ES40 and to the AlphaServer GS80 series, asC >   these have newer bus (well, actually, crossbar) system designs.  > F > :Also, I now see some #1 rated ECOs that are to be applied to all GS systems.& > :Does that include my upgraded 8400? > I >   Though without knowledge of the particular ECO kits are referenced, I - >   would tend to assume the answer is "yes".  > G > :I have not had a Compaq (or Digital) sales person call me for years.  > G >   Though not a "Sales Puke" :-), please send me the requisite contact H >   information (offline), and I'll work on getting this effort rolling. >  >   Hoff >   ("Engineering Puke" :-)  > , >  --------------------------- pure personal# opinion --------------------------- 1 >    Hoff (Stephen) Hoffman   OpenVMS Engineering  hoffman#xdelta.zko.dec.com >    ------------------------------  % Date: Fri, 28 Jul 2000 11:36:06 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com> & Subject: Re: Apache for OpenVMS (VAX?)) Message-ID: <39815406.48F544C2@gtech.com>    FAZEKAS Mihaly wrote: ) > I'm searching Apache for OpenVMS/Alpha. C > But i want VAX version. Can i find OpenVMS/VAX version of Apache? ' > I'm checked in the web, but not find.    I think Apache is Alpha only.    > What can i do?  - There are plenty of Web-servers for VMS VAX !   7 OSU and WASD was the two obvious ones to choose among !    Arne   ------------------------------  % Date: Fri, 28 Jul 2000 08:46:46 +0100 ' From: Hitendra Patel <h.patel@rl.ac.uk>  Subject: Re: BACKUP COMMAND ( Message-ID: <39813A66.D17EF010@rl.ac.uk>  	 Dear All,   I Well thank you all for replying to my question. Unfortunately the exabyte F drive gave up but I need manage to pinch one and completed the restore successfully for a user.   Once again  'THANK YOU'      Hitendra Patel   ------------------------------  # Date: Fri, 28 Jul 2000 16:56:30 GMT ( From: "Randall Becker" <nospam@home.com>- Subject: Re: CETS2000 - Very Important Notice ? Message-ID: <2Xig5.17893$07.792517@news1.rdc1.on.wave.home.com>   F Did anyone see any NSK topics on that list? Anyone? ServerNet? Sierra?   Anyone?   B "Information CETS2000" <Information@CETS2000.com> wrote in message, news:8lqhl0$7bv$1@slb0.atl.mindspring.net... > August 1, 2000 >  > Dear Fellow DECUS Member:  > ; > By now you should be aware of Compaq Enterprise Technical ; > Symposium 2000, which DECUS is co-producing along with NT A > Wizards, Compaq's ASE program, and Compaq's CSA program.  There = > is an important new feature for DECUS members, of which you : > should be aware.  The very valuable Hands-on Interactive@ > Workshops require advanced reservation prior to the Symposium.C > This advance reservation will begin in late August.  All of these B > Workshops have very limited space available and the seats in the0 > most popular workshops will disappear quickly. > > > What are Hands-on Interactive Workshops?  They are your bestA > opportunity to "learn by doing," because experience is the best > > teacher.  These small, instructor-led workshops give you theC > opportunity to work with some of Compaq's most advanced platforms @ > and products.  For example, you can use the lab environment to? > learn about installing fibre channel SANs, configuring remote ? > server management, planning for disaster recovery, setting up @ > security, optimizing server clusters, and so much more.  TheseC > workshops give you the most practical technical education you can = > get.  There are dozens of workshops where you can hone your B > skills (see "Consideration" list below).  Truckloads of servers,C > PCs, workstations, storage units, backup systems - you name it! - > > are being hauled to Los Angeles for the Hands-on Interactive > Workshops. > @ > After you register for the symposium, you will be sent special> > instructions for choosing and registering for your preferredB > Hands-on Interactive Workshops.  I strongly suggest you consider@ > completing your registration process prior to the last week in= > August in order to have the best opportunity to reserve the A > Workshops you want.  Other groups involved in Compaq Enterprise @ > Technical Symposium 2000 are experienced with this process andA > know to register early to be immediately eligible when the this A > special registration process opens.  Seats go first come, first 	 > served.  > ? > Note: To insure fairness each attendee will be limited to one C > advanced registration, so choose wisely.  You may attend a second = > workshop by registering on-site if space becomes available.  > > > For questions about CETS-2000 visit www.CETS2000.com or send# > email to Information@CETS2000.com  >  >  >  > Kristi Browder, > U.S. DECUS Executive Sponsor for CETS 2000 > Kristi.Browder@CETS2000.com  >  >  >  > 5 >  Hands-on Interactive Workshops Under Consideration  > > > The following is a working list showing some of the Hands-onB > Interactive Workshops being considered for the Compaq EnterpriseA > Technical Symposium 2000.  This is not a final list and changes @ > are likely.  There will be deletions and additions.  The finalB > list will be available in late August. However, these topics areC > representative of the types of Hands-on Interactive Workshops you  > can expect at the symposium. > < > Installing & Configuring AlphaServers for ProLiant JunkiesD > Building and Installing a Web Infrastructure with Apache/PHP/MySQL1 > Getting the Most Out of Alpha Performance Tools " > Building a Linux Beowulf ClusterC > Compaq Enterprise Toolkit - The OpenVMS Edition Hands-On Workshop A > Configuring an Ultra - High Availability Environment on OpenVMS ) > UNIX - Get More out of Your Application  > Installing Linux on Alpha 3 > ProLiant/Linux Web Clusters in a DISA Environment 2 > Implementing Intelligent Manageability Solutions. > Thin Client Configuration and Management Lab% > TaskSmart C - General Configuration & > Tools for Deploying ProLiant Servers@ > Compaq Insight Manager XE 2.0 - Basic Implementation and Usage > SCO Clustering Workshop $ > Novell NetWare Clustering WorkshopC > Compaq Insight Manager XE 2.0 - Advanced Implementation and Usage ? > Configuring Compaq Enterprise Management Tools for Management  > Service Providers " > Windows 2000 Clustering Workshop) > Tasksmart C Series, Layer 4 and Layer 7 # > SANworks Storage Resource Manager $ > Compaq SANworks Virtual Replicator( > RA4X00 and EBS SAN Technology Workshop) > Enterprise Backup Solution Architecture 2 > Multi Vendor RA8000FC Installation on NT and SUN6 > Multi Vendor RA8000FC Installation on TRU64 and OVMS8 > Multi Vendor RA8000FC Installation on Novell and Linux+ > Compaq Performance Analysis Tool Workshop 2 > Linux Installation/Configuration on StorageWorks8 > Windows 2000 Administration for the NT 4 Administrator > LINUX for the VMS User > LINUX GUI / > Building a Web Page with Microsoft Front Page . > Disk Cloning Technology for IT Professionals > Intro to HTML 3 > Intro to SDA by Viewing OpenVMS Process Internals 7 > Tru64/TruCluster Administration Intro for VMS Cluster  > Administrators# > VMS Cluster Transitions in Action  > VMS Lock Manager in Action > RMS File Repair Techniques >  >  >    ------------------------------    Date: 28 Jul 2000 10:14:45 -07001 From: nothome@spammers.are.scum (Malcolm Dunnett) D Subject: Datatrieve (was Re : Interbase 6.0 released as Open Source), Message-ID: <IbJ76BR44wKA@malvm2.mala.bc.ca>  9 In article <009EDBB5.BDE375E0@SSRL04.SLAC.STANFORD.EDU>,  U        winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes:  >  > Q > I'm actually using Datatrieve very happily now, mostly in batch mode to produce P > reports and web pages for automatic distribution.  Runs well, easy to code in,N > and on an AS2100 5/300 it isn't too bad to run for interactive use.  On the Q > VAX 11/750 I learned it on, I spent a lot of time figuring out how to optimize.  > O       We have some applications here which use Datatrieve quite extensively and N are also quite happy with it ( on a 500MHz EV5 ). Getting rid of the CDD cruftO and going to text based dictionaries eliminated most of my complaints with DTR.   7      If only they'd make it work with Oracle tables :-)   M ============================================================================= M Malcolm Dunnett      Malaspina University-College   Email: dunnett@mala.bc.ca H Information Systems  Nanaimo, B.C. CANADA V9R 5S5     Tel: (250)755-8738   ------------------------------  % Date: Fri, 28 Jul 2000 02:23:31 -0400 2 From: rdeininger@mindspring.com (Robert Deininger)2 Subject: Re: Exabyte Mammoth-2 tape drives and VMSL Message-ID: <rdeininger-2807000223320001@user-2ivecal.dialup.mindspring.com>  O In article <8lq1kg$kf$1@husk.cso.niu.edu>, system@niuhep.physics.niu.edu wrote:    > Hello, > H > It looks like our collaborators are going with  Exabyte Mammoth-2 tapeC > drives, so we will also, any advice, known gotchas, etc. would be  > appreciated.    N I haven't used any of the recent exabytes, but the old ones stank.  OverpricedI tape-shredders.  Slow as cold molasses.  Torture devices.  I suspect they K named the latest generation Mammoths because they are (please God) about to  go extinct.   J Have they improved a whole lot?  If not, you're going to wish you'd boughtK DLTs.  You should try to turn your collaborators around.  But collaborators ' can be very similar to flocks of sheep.   M It's late, and I'm tired, or I'd tell you what I really think about exabites.    --   Robert Deininger rdeininger@mindspring.com    ------------------------------  % Date: Fri, 28 Jul 2000 11:29:32 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com> 2 Subject: Re: Exabyte Mammoth-2 tape drives and VMS) Message-ID: <3981527C.50BC421A@gtech.com>   $ system@niuhep.physics.niu.edu wrote:H > It looks like our collaborators are going with  Exabyte Mammoth-2 tapeC > drives, so we will also, any advice, known gotchas, etc. would be  > appreciated.   8mm technology sucks.    Go for DLT instead.    Arne   ------------------------------  % Date: Fri, 28 Jul 2000 16:50:45 +0100 - From: Tim Llewellyn <tim.llewellyn@bbc.co.uk> 2 Subject: Re: Exabyte Mammoth-2 tape drives and VMS) Message-ID: <3981ABD5.8426DE51@bbc.co.uk>    Arne Vajhj wrote:  & > system@niuhep.physics.niu.edu wrote:J > > It looks like our collaborators are going with  Exabyte Mammoth-2 tapeE > > drives, so we will also, any advice, known gotchas, etc. would be0 > > appreciated. >  > 8mm technology sucks.A >  > Go for DLT instead.e  G This does not help if he needs to exchange data with his collaborators.a  G Academic cheapskate project by the sound of it, I share Arne's distrusttG of exabyte drives.  Back in the early 90's they were GIVING drives awayMK at CERN to anybody from a collaborating instuitute who would haul one away, G they got so p***ed with calling in the engineer to reallign the drives.r   --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.ukt  A I speak for myself only and my views in no way represent those ofe MedAS or the BBC.    ------------------------------  % Date: Fri, 28 Jul 2000 20:27:42 +0010p% From: paddy.o'brien@zzz.tg.nsw.gov.aue Subject: Re: Fortune cookies?n5 Message-ID: <01JSBAODBLXE002B0K@tgmail.tg.nsw.gov.au>t   Martin,/  N >> Any fortune cookie or "quote of the day" is remarkably easy to write in anyJ >> language.  The major problem is in increasing your "quotes" database.   >wB >	Yup, and the good thing about this version is that it comes with >	quite a number of quotes ;-) > 	 >> I haveeQ >> lost track of where colleagues picked up quotes from.  I know that one source   is+ >> Art Ragosta's database of pithy sayings.n >> eN >> If anyone needs a copy of what we have collected, email me because I do notP >> think any are copyright.  Pithy comments from Art and colleagues, and others,M >> and newspaper headlines, some added from .au papers that are as bad as anye+ >> others.  (Art, do you have a copyright?)a >2> >	Definitly interested, can you make it available for download >	somewhere?  C Starting from the back, I do not have access to any download areas.S  Q I have had private "swaps" with a couple of people.  Martin Vorlaender and Terry  P Kennedy.  My program is very primitive and I'm looking at Terry's, though on my A ES40 it takes very little time to count and then randomly select.M  Q As a sidebar, although I said that we included Art Ragosta's, my colleague tells cP me that at present we haven't.  We have picked up his gems, but they need a bit Q of work to put them into the form that we analyse.  A commodity like time is not r always readily available :-)  O Since I have now mailed to a few people who do have sites, I am sure that they  = can probably incorporate with their collection on their site.e   Regards, Paddy   Paddy O'Brien, Transmission Development,i
 TransGrid, PO Box A1000, Sydney South,  NSW 2000, Australiap   Tel:   +61 2 9284-3063 Fax:   +61 2 9284-3050& Email: paddy.o'brien@zzz.tg.nsw.gov.au  M Either "\'" or "\s" (to escape the apostrophe) seems to work for most people,n; but that little whizz-bang apostrophe gives me little spam.    ------------------------------  % Date: Fri, 28 Jul 2000 19:07:46 +0200-= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>0 Subject: GTECH DK hiring) Message-ID: <3981BDE2.D831EA99@gtech.com>j  8 GTECH DK needs to hire at least one developer plus a QA.  . So anyone familiar with some of the following:   VMS [very important]   Unix (Tru64 & Solaris)   OLTP   SQL databases (Sybase)   e-business)   C [only relevant for developer postion] ,   Java [only relevant for developer postion]6 and already living in Denmark or willing to relocate -3 please feel free to contact me for more info if you  are interested.n  0 And yes - GTECK DK is a nice place to work ! :-)   Arne   ------------------------------  # Date: Fri, 28 Jul 2000 07:42:26 GMTr From: tgreenwood@my-deja.com* Subject: help with MACRO compiler warnings( Message-ID: <8lrdh1$ob$1@nnrp1.deja.com>   Hi  A I'm using MACRO as part of a script to compile a table containingcA details of packages which will be accessed from a layered productoE (DSM), but I think the script is not working because it pre-dates oureF move to Alpha (a near-identical script worked okay last time we played this game on a Vax).  B When we compile the table we get lots of 'global symbol unaligned'> warnings which as far as I can tell are relating to a possible, performance problem which doesn't bother us.  D By fiddling with the command line qualifiers for MACRO I've made theF warnings stop and the compile now completes, but when I try to use the@ package from DSM it objects because the image 'contains compiler
 warnings'.  E Given that they're just warnings is there any way I can tell MACRO toe4 not leave any record of them so I can use the image?   Cheers Tony.c   -- Tony Greenwood Logabex Ltd. Rowlands Castle, UK2    & Sent via Deja.com http://www.deja.com/ Before you buy.a   ------------------------------  # Date: Fri, 28 Jul 2000 10:22:42 GMT1= From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-)a. Subject: Re: help with MACRO compiler warnings0 Message-ID: <009EDBE4.B363000C@SendSpamHere.ORG>  G In article <8lrdh1$ob$1@nnrp1.deja.com>, tgreenwood@my-deja.com writes:s >Hin >iB >I'm using MACRO as part of a script to compile a table containingB >details of packages which will be accessed from a layered productF >(DSM), but I think the script is not working because it pre-dates ourG >move to Alpha (a near-identical script worked okay last time we playeda >this game on a Vax).u > C >When we compile the table we get lots of 'global symbol unaligned' ? >warnings which as far as I can tell are relating to a possibleh- >performance problem which doesn't bother us.l  F Can you post the exact message?  As far as I can tell, the message youG are getting is: %AMAC-I-UNALGLOLAB, unaligned global label, and this isn" an informational -- not a warning.    E >By fiddling with the command line qualifiers for MACRO I've made theoG >warnings stop and the compile now completes, but when I try to use theoA >package from DSM it objects because the image 'contains compilerh >warnings'.r  @ Then there is something besides the 'unaligned' being generated.  F >Given that they're just warnings is there any way I can tell MACRO to5 >not leave any record of them so I can use the image?t   Please post the "warnings".O   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COM    ------------------------------  % Date: Fri, 28 Jul 2000 20:48:44 +0010u% From: paddy.o'brien@zzz.tg.nsw.gov.aum. Subject: Re: help with MACRO compiler warnings5 Message-ID: <01JSBBEG8CQQ002BFB@tgmail.tg.nsw.gov.au>   F >Given that they're just warnings is there any way I can tell MACRO to5 >not leave any record of them so I can use the image?   P A repeat of the warnings (rather the fact that they existed in some .OBJ) would Q only appear at link time.  This would not obviate the creation of an executable, e nor be generated at runtime.  F [I did not see the original, just picked this up from VAXMAN's reply.]   Regards, Paddy   Paddy O'Brien, Transmission Development,'
 TransGrid, PO Box A1000, Sydney South,  NSW 2000, Australian   Tel:   +61 2 9284-3063 Fax:   +61 2 9284-3050& Email: paddy.o'brien@zzz.tg.nsw.gov.au  M Either "\'" or "\s" (to escape the apostrophe) seems to work for most people,v; but that little whizz-bang apostrophe gives me little spam.o   ------------------------------  # Date: Fri, 28 Jul 2000 11:34:27 GMT2 From: tgreenwood@my-deja.com. Subject: Re: help with MACRO compiler warnings) Message-ID: <8lrr43$9k1$1@nnrp1.deja.com>.  0 In article <009EDBE4.B363000C@SendSpamHere.ORG>,    system@SendSpamHere.ORG wrote:  H > Can you post the exact message?  As far as I can tell, the message youF > are getting is: %AMAC-I-UNALGLOLAB, unaligned global label, and this is$ > an informational -- not a warning. >-  # You're right - it's a whole slew of:* %AMAC-I-UNALGLOLAB, unaligned global labelC and you're quite right, it *is* an informational, but it's the only ; message that I'm getting during the compile/link procedure.n  : When I attempt to access the final image from DSM I get...8 %DSM-E-ZCSIGNAL, condition signalled by external routine' -DSM-F-IMAGEACT, Error activating imageb@ -LIB-W-EOMWARN, compilation warnings in module DSM$ECALL_BANKWIZ  E It is of course possible that the 'compilation' mentioned here is noteE what I'm thinking of but is something that's happening when DSM tries  to run the image.   C Is information like 'compilation warnings' stored in images or am Ie. barking up the wrong tree entirely (as usual)?   Thanks Tony.e    & Sent via Deja.com http://www.deja.com/ Before you buy.d   ------------------------------  # Date: Fri, 28 Jul 2000 12:20:02 GMTi= From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-) . Subject: Re: help with MACRO compiler warnings0 Message-ID: <009EDBF5.173A4CEB@SendSpamHere.ORG>  H In article <8lrr43$9k1$1@nnrp1.deja.com>, tgreenwood@my-deja.com writes:1 >In article <009EDBE4.B363000C@SendSpamHere.ORG>, ! >  system@SendSpamHere.ORG wrote:e >oI >> Can you post the exact message?  As far as I can tell, the message younG >> are getting is: %AMAC-I-UNALGLOLAB, unaligned global label, and thism >iso% >> an informational -- not a warning.l >> >c$ >You're right - it's a whole slew of+ >%AMAC-I-UNALGLOLAB, unaligned global label D >and you're quite right, it *is* an informational, but it's the only< >message that I'm getting during the compile/link procedure. > ; >When I attempt to access the final image from DSM I get...u9 >%DSM-E-ZCSIGNAL, condition signalled by external routineo( >-DSM-F-IMAGEACT, Error activating imageA >-LIB-W-EOMWARN, compilation warnings in module DSM$ECALL_BANKWIZn >eF >It is of course possible that the 'compilation' mentioned here is notF >what I'm thinking of but is something that's happening when DSM tries >to run the image. >rD >Is information like 'compilation warnings' stored in images or am I/ >barking up the wrong tree entirely (as usual)?p >r >Thanksa >Tony. >  >u' >Sent via Deja.com http://www.deja.com/y >Before you buy.  H Try $ MACRO/WARN=NOINFO and see if there are any messages you might have overlooked.    --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COMe   ------------------------------   Date: 28 Jul 2000 13:25:08 GMT2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman). Subject: Re: help with MACRO compiler warnings6 Message-ID: <8ls1jk$rdd$1@mailint03.im.hou.compaq.com>  G In article <8lrdh1$ob$1@nnrp1.deja.com>, tgreenwood@my-deja.com writes:lB :I'm using MACRO as part of a script to compile a table containingB :details of packages which will be accessed from a layered product	 :(DSM)....  ?   A short code example of the problem and the error(s), please?.B   (Please also include the OpenVMS version and the MACRO command.)  B   Would this be a shareable image transfer vector module?  (If so,C   please see the Shareable Image Cookbook referenced in the OpenVMSpD   FAQ, as details of transfer vectors differ between VAX and Alpha.)  N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------  # Date: Fri, 28 Jul 2000 14:12:45 GMTs From: tgreenwood@my-deja.com. Subject: Re: help with MACRO compiler warnings) Message-ID: <8ls4cn$gab$1@nnrp1.deja.com>t  0 In article <009EDBF5.173A4CEB@SendSpamHere.ORG>,    system@SendSpamHere.ORG wrote:E > Try $ MACRO/WARN=NOINFO and see if there are any messages you might  have
 > overlooked.   ? No messages from the MACRO command, just %LINK-W-MULDEF, symbol 1 decc$unlink multiply defined from the link phase.i    , In case it helps, this is the code involved:   RECOMPILE.COM:9 $	INSTALL REMOVE SYS$SYSDEVICE:[BANKWIZ]DSM$ECALL_BANKWIZ ( $	DEFINE LNK$LIBRARY SYS$LIBRARY:VAXCRTL $! $! Rebuild DSM$ECALL_BANKWIZ $!' $	MACRO BANKWIZ:DSM$ECALL_BANKWIZ_TABLEe& $	LINK/SHARE=BANKWIZ:DSM$ECALL_BANKWIZH BANKWIZ:DSM$ECALL_BANKWIZ_TABLE.OBJ,SYS$LIBRARY:DSM$ECALL.OPT/OPTION,BAN KWIZ:OBJECT.OLB/LIBRARYo $! $! Rebuild DSM$ECALL $!' $	MACRO BANKWIZ:DSM$ECALL_BANKWIZ_PKTABf  $	LINK/SHARE=BANKWIZ:DSM$ECALL -  A 	BANKWIZ:DSM$ECALL_BANKWIZ_PKTAB,sys$library:dsm$ecall.opt/optionr $! $! Copy to SYS$SHARE $!A $	COPY/LOG BANKWIZ:DSM$ECALL.EXE SYS$COMMON:[SYSLIB]DSM$ECALL.EXE < $	COPY/LOG BANKWIZ:DSM$ECALL_BANKWIZ.EXE SYS$COMMON:[SYSLIB] DSM$ECALL_BANKWIZ.EXEp" $	INSTALL REPL SYS$COMMON:[SYSLIB]# DSM$ECALL.EXE/OPEN/HEAD/WRITE/SHAREo" $	INSTALL REPL SYS$COMMON:[SYSLIB]+ DSM$ECALL_BANKWIZ.EXE/OPEN/HEAD/WRITE/SHAREd   ======================   DSM$ECALL_BANKWIZ_PKTAB.MAR:   		.TITLE		DSM$ECALLW 		.IDENT		/DSM V7.1/    $ 		.LIBRARY	/SYS$LIBRARY:DSM$ECMACRO/  4 		PACKAGE_INIT	; Init Package Table  - Must be first	 statement     + 		PACKAGE_DEFINE -		; define "ZDSM" package <                         PACKAGE_NAME=ZDSM,-	; as package # 0 			IMAGE_NAME=DSM$SHAREH  , 		PACKAGE_DEFINE - 		; define "ZLIB" package' 			PACKAGE_NAME=ZLIB,-	; as package # 1t 			IMAGE_NAME=DSM$SHARE   & 		PACKAGE_DEFINE	-		; define "EXAMPLE" packaget* 			PACKAGE_NAME=BANKWIZ,-	; as package # 2 			IMAGE_NAME=DSM$ECALL_BANKWIZ    		PACKAGE_FINISH 		.END   ========================   DSM$ECALL_BANKWIZ_TABLE.MARm  ' 	.LIBRARY /SYS$LIBRARY:DSM$ECMACRO.MLB/D   	ECALLINI		; Initialize Packagee   ; INITSCANSS. ROUTINE		CALLNAME=INITSCANS	LINKNAME=INITSCANS RETURN		VALUE			TYPE=LONG5$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG   ; DENITSCANS0 ROUTINE		CALLNAME=DENITSCANS	LINKNAME=DENITSCANS RETURN		VALUE			TYPE=LONG    ; CHECKSCANS0 ROUTINE		CALLNAME=CHECKSCANS	LINKNAME=CHECKSCANS RETURN		VALUE			TYPE=LONGo% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZu$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZn$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ'$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG     ; VERSIONSCANS4 ROUTINE		CALLNAME=VERSIONSCANS	LINKNAME=VERSIONSCANS RETURN		VALUE			TYPE=LONGy0 ARGUMENT        MECHANISM=POINTER     TYPE=ASCIZ   ; SPLITSCANS0 ROUTINE		CALLNAME=SPLITSCANS	LINKNAME=SPLITSCANS RETURN		VALUE			TYPE=LONGd% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZe$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZe$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZk$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZr$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZe$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZv$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG   	.ENDb =====================     & Sent via Deja.com http://www.deja.com/ Before you buy.t   ------------------------------  # Date: Fri, 28 Jul 2000 14:41:27 GMTr= From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-)i. Subject: Re: help with MACRO compiler warnings0 Message-ID: <009EDC08.D88DF7C7@SendSpamHere.ORG>  H In article <8ls4cn$gab$1@nnrp1.deja.com>, tgreenwood@my-deja.com writes:1 >In article <009EDBF5.173A4CEB@SendSpamHere.ORG>,:! >  system@SendSpamHere.ORG wrote:-F >> Try $ MACRO/WARN=NOINFO and see if there are any messages you might >haver >> overlooked. >g@ >No messages from the MACRO command, just %LINK-W-MULDEF, symbol2 >decc$unlink multiply defined from the link phase. >  >h- >In case it helps, this is the code involved:a     Care to post the macros?   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COMr   ------------------------------  # Date: Fri, 28 Jul 2000 14:57:17 GMT0 From: tgreenwood@my-deja.com. Subject: Re: help with MACRO compiler warnings) Message-ID: <8ls709$ilb$1@nnrp1.deja.com>8  6 In article <8ls1jk$rdd$1@mailint03.im.hou.compaq.com>,&   hoffman@xdelta.zko.dec.nospam wrote:  A >   A short code example of the problem and the error(s), please? D >   (Please also include the OpenVMS version and the MACRO command.) > D It's OpenVMS 7.1 - the (not very short) code is tagged on the end of
 this message.     D >   Would this be a shareable image transfer vector module?  (If so,E >   please see the Shareable Image Cookbook referenced in the OpenVMSaF >   FAQ, as details of transfer vectors differ between VAX and Alpha.) > 9 That whooshing noise is your question going over my head!n@ If I understand your meaning then yes I think it is, but I don'tF *think* it's an issue. You may of course know better in which case I'd
 like to know!t   Cheers Tony.d   The code involved is:m   RECOMPILE.COM:9 $	INSTALL REMOVE SYS$SYSDEVICE:[BANKWIZ]DSM$ECALL_BANKWIZo( $	DEFINE LNK$LIBRARY SYS$LIBRARY:VAXCRTL $! $! Rebuild DSM$ECALL_BANKWIZ $!' $	MACRO BANKWIZ:DSM$ECALL_BANKWIZ_TABLEr& $	LINK/SHARE=BANKWIZ:DSM$ECALL_BANKWIZH BANKWIZ:DSM$ECALL_BANKWIZ_TABLE.OBJ,SYS$LIBRARY:DSM$ECALL.OPT/OPTION,BAN KWIZ:OBJECT.OLB/LIBRARYa $! $! Rebuild DSM$ECALL $!' $	MACRO BANKWIZ:DSM$ECALL_BANKWIZ_PKTABe  $	LINK/SHARE=BANKWIZ:DSM$ECALL -  A 	BANKWIZ:DSM$ECALL_BANKWIZ_PKTAB,sys$library:dsm$ecall.opt/option" $! $! Copy to SYS$SHARE $!A $	COPY/LOG BANKWIZ:DSM$ECALL.EXE SYS$COMMON:[SYSLIB]DSM$ECALL.EXEu< $	COPY/LOG BANKWIZ:DSM$ECALL_BANKWIZ.EXE SYS$COMMON:[SYSLIB] DSM$ECALL_BANKWIZ.EXEs" $	INSTALL REPL SYS$COMMON:[SYSLIB]# DSM$ECALL.EXE/OPEN/HEAD/WRITE/SHAREs" $	INSTALL REPL SYS$COMMON:[SYSLIB]+ DSM$ECALL_BANKWIZ.EXE/OPEN/HEAD/WRITE/SHAREa   ======================   DSM$ECALL_BANKWIZ_PKTAB.MAR:   		.TITLE		DSM$ECALL, 		.IDENT		/DSM V7.1/    $ 		.LIBRARY	/SYS$LIBRARY:DSM$ECMACRO/  4 		PACKAGE_INIT	; Init Package Table  - Must be first	 statementi    + 		PACKAGE_DEFINE -		; define "ZDSM" packagei<                         PACKAGE_NAME=ZDSM,-	; as package # 0 			IMAGE_NAME=DSM$SHAREt  , 		PACKAGE_DEFINE - 		; define "ZLIB" package' 			PACKAGE_NAME=ZLIB,-	; as package # 1l 			IMAGE_NAME=DSM$SHAREd  & 		PACKAGE_DEFINE	-		; define "EXAMPLE" package * 			PACKAGE_NAME=BANKWIZ,-	; as package # 2 			IMAGE_NAME=DSM$ECALL_BANKWIZo   		PACKAGE_FINISH 		.END   ========================   DSM$ECALL_BANKWIZ_TABLE.MARg  ' 	.LIBRARY /SYS$LIBRARY:DSM$ECMACRO.MLB/    	ECALLINI		; Initialize Packagey   ; INITSCANSf. ROUTINE		CALLNAME=INITSCANS	LINKNAME=INITSCANS RETURN		VALUE			TYPE=LONGn$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG   ; DENITSCANS0 ROUTINE		CALLNAME=DENITSCANS	LINKNAME=DENITSCANS RETURN		VALUE			TYPE=LONGy   ; CHECKSCANS0 ROUTINE		CALLNAME=CHECKSCANS	LINKNAME=CHECKSCANS RETURN		VALUE			TYPE=LONGm% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZx$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZs$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ $ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ-$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZj$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ.$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG     ; VERSIONSCANS4 ROUTINE		CALLNAME=VERSIONSCANS	LINKNAME=VERSIONSCANS RETURN		VALUE			TYPE=LONG 0 ARGUMENT        MECHANISM=POINTER     TYPE=ASCIZ   ; SPLITSCANS0 ROUTINE		CALLNAME=SPLITSCANS	LINKNAME=SPLITSCANS RETURN		VALUE			TYPE=LONGg% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZt$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZe$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZ2$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZC$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZs$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZa$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG% ARGUMENT	MECHANISM=POINTER	TYPE=ASCIZr$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG$ ARGUMENT	MECHANISM=POINTER	TYPE=LONG   	.ENDg =====================w        & Sent via Deja.com http://www.deja.com/ Before you buy.n   ------------------------------    Date: 28 Jul 2000 16:02:08 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>u2 Subject: Re: I/O Size and I/Os per second overviewH Message-ID: <y4hf9ajqu7.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  , young_r@eisner.decus.org (Rob Young) writes:  @ > http://www.dell.com/us/en/k12/topics/vectors_1999-fcraidpe.htm0 >                                Data Throughput [...] E > a 64-KB data transfer may [...] require 16 ms to transfer the data.n  G If I do my calculations right, that means a transfer rate of 60*64 KB/sa or less than 4 MB/s.  / In what decade did they write that - the 1970s?1   	Jan   ------------------------------    Date: 28 Jul 2000 11:29:04 -0500* From: young_r@eisner.decus.org (Rob Young)2 Subject: Re: I/O Size and I/Os per second overview+ Message-ID: <YRZ6nhNFzIYu@eisner.decus.org>y   In article <y4hf9ajqu7.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>, Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> writes:. > young_r@eisner.decus.org (Rob Young) writes: > A >> http://www.dell.com/us/en/k12/topics/vectors_1999-fcraidpe.html1 >>                                Data Throughpute > [...])F >> a 64-KB data transfer may [...] require 16 ms to transfer the data. > I > If I do my calculations right, that means a transfer rate of 60*64 KB/sm > or less than 4 MB/s. > 1 > In what decade did they write that - the 1970s?a >   E 	You find these guides, and misinterpret (fairly recent guides, usinge? 	"close" to state of the art, right?) and not quite up to speed 7 	on all this... ZING!  Same metrics as Dell found here:R  > http://oradoc.photo.net/ora8doc/DOC/server803/A54638_01/io.htm  C 	Going back to that formula, that Dell number has to be better thane 	a plain old SCSI disk:r  J         Maximum IOPS per disk drive = 1/(seek + rotation + transfer times)G                                 = 1/(8 ms + 0.5/7200 RPM + 64KB/6 MBps) D                                 = 1/(8 + 4.2 + 10.4) = 1/(22.6 msec))                                 = 44 IOPS   t  A 	Maybe an IDE at 7200 RPM (Bill pointed out IDE does 10-20 MB/sec- 	today):  J         Maximum IOPS per disk drive = 1/(seek + rotation + transfer times)H                                 = 1/(8 ms + 0.5/7200 RPM + 64KB/12 MBps)C                                 = 1/(8 + 4.2 + 5.2) = 1/(17.4 msec)u)                                 = 57 IOPSR  ; 	They are using 7200 RPM IDE drives as their baseline...(or  	something similar :-).(   	What a DRAG ...   				Robe   ------------------------------    Date: 28 Jul 2000 02:49:48 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: I/Os per second, how do you increase them?s+ Message-ID: <J+LLM+sLeNVo@eisner.decus.org>a  R In article <8lr1bp$1n0$1@pyrite.mv.net>, "Bill Todd" <billtodd@foo.mv.com> writes: > 7 > Rob Young <young_r@eisner.decus.org> wrote in message ' > news:kwzvSbaUAGhW@eisner.decus.org...iM >> In article <8lq91r$8gp$1@pyrite.mv.net>, "Bill Todd" <billtodd@foo.mv.com> 	 > writes:s >> >> >L >> > You missed the fore-going context:  35 ms. per 64 KB (sequential, but I > let-J >> > that go for the moment) I/O was the figure *Rob* was quoting from theJ >> > on-line source to which he provided a pointer.  I was simply pointing > outpH >> > that if one accepted that figure (which one should not - it clearly	 > doesn'tr; >> > apply to the matters under discussion here, nor to anyI > competently-designedL >> > access mechanism anywhere else that uses modern disks) as the cost of a > 64L >> > KB access then, given contemporary disk characteristics, the disk could > not-J >> > (even satisfying smaller requests) exceed about 32 I/O operations per > second0 >> > (that's called reductio ad absurdum, IIRC). >> > >>@ >> That chart shows 35 ms for 127 KB transfer.  A 64 KB transfer> >> takes 16 ms as the Dell reference points out and the OracleE >> reference shows.  My first post contained an error as you spotted.oB >> Divide by two error.  127 blocks makes for a 64 KB I/O , double8 >> the number of I/Os per second but halve the I/O size. > G > Nope - you still have no clue, and won't until you figure out how theeN > numbers relate to each other.  For a start, the number of I/Os per second isK > nowhere nearly inversely proportional to the I/O size until the I/O sizesv > reach megabytes in size. >   : 	Uhh.. not quite.  Doubling the I/O size according to that6 	chart doubles the transfer time... The chart reflects; 	older disks and does make it lean more towards an inverse.h  M > You can also ignore the Dell numbers (which were specifically offered as anaK > *example*, but happen to be close to an order of magnitude lower than theiK > transfer rates of current disks):  a 64 KB transfer takes between 2 and 3e! > ms. on modern disks, not 16 ms.s >   " 	The formula to calculate IOPS is:  J         Maximum IOPS per disk drive = 1/(seek + rotation + transfer times)  2 	Plugging in some values for an older disk (SCSI):  G                                 = 1/(8 ms + 0.5/7200 RPM + 16KB/6 MBps) C                                 = 1/(8 + 4.2 + 2.7) = 1/(14.9 msec)h)                                 = 67 IOPSe  = 	67 IOPS for 16 KB I/O size on an older disk.  The numbers in > 	the Oracle chart look like older disks.  I suspect John Nixon9 	and many of us have RZ29 disks, he is working with HSJs,a 	7200 RPM SCSI disks.m  9 	Looking at the Seagate 15K disk due out later this year:   R http://www.seagate.com/cda/products/discsales/enterprise/family/0,1130,246,00.html  1 	IOPS 	= 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms)u8 		= 160 IOPS  (I read 140, so I am off here but there is( 				command overhead so not a one to one 				correspondence)l  ( 	A little something on command overhead:    K IOPS Delivery - The theoretical random IOPS that a typical high performanceNL SCSI disk drive can deliver is the sum of an average seek (9 ms.), a latencyH period (4 ms.), and a 1 ms. delay that takes into account transfer time,N command handshaking and decode. Therefore, you have 9+4+1=14 ms. per transfer.M 1000 ms./14 ms.=71.43 IOPS per disk. This number assumes that the transfer isiN 8K or less. Larger I/O's will result in a reduced IOPS number, smaller ones doN not go faster since 95% of the overhead is either mechanical or command time.    	64 KB I/O:   4 	IOPS 	= 1/(3.9ms + 2.1ms + 64KB/42.5MBps = 1.47 ms) 		= 134 IOPS  B 	Drop off is noticeable but much better IOPS with larger transfers 	on newer disks.   	By the way, your claim:    K > nowhere nearly inversely proportional to the I/O size until the I/O sizess > reach megabytes in size.    4 	IOPS 	= 1/(3.9ms + 2.1ms + .5MB/42.5MBps = 11.8 ms) 		= 56 IOPSe  3 	IOPS 	= 1/(3.9ms + 2.1ms + 1MB/42.5MBps = 23.5 ms)  		= 34 IOPSr  1 	IOPS 	= 1/(3.9ms + 2.1ms + 2MB/42.5MBps = 47 ms)- 		= 19 IOPSn  $ 	Close to inverse from 1 MB to 2 MB.  > 	And Bill, it isn't always about being clueless or overlookingG 	or not paying attention or whatever.  Sometimes it is about missing a :F 	piece or in this case the formula to calculate all this.  Between us ? 	we could have probably looked at that Oracle chart and said:  nA 	"oh, those must be 7200 RPM SCSI drives" and also concluded theyrC 	couldn't get more raw or block I/O out of them if the Queen MotheraG 	wanted it.   I couldn't have done that yesterday, I can do that today.d   				RobK   ------------------------------    Date: 28 Jul 2000 03:14:28 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: I/Os per second, how do you increase them?[+ Message-ID: <bqPAaO6oP1cC@eisner.decus.org>A  X In article <J+LLM+sLeNVo@eisner.decus.org>, young_r@eisner.decus.org (Rob Young) writes:   > ; > 	Looking at the Seagate 15K disk due out later this year:L > T > http://www.seagate.com/cda/products/discsales/enterprise/family/0,1130,246,00.html > 3 > 	IOPS 	= 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms)_: > 		= 160 IOPS  (I read 140, so I am off here but there is* > 				command overhead so not a one to one > 				correspondence)S >    	One last swipe for Deja sake:  ' 	Adding 1 ms for command overhead = CO:   ;  	IOPS 	= 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms + 1ms CO)S>  		= 142 IOPS  (formerly 160, there is the 140 we read about).    	64 KB I/O:  R>  	IOPS 	= 1/(3.9ms + 2.1ms + 64KB/42.5MBps = 1.47 ms + 1ms CO)  		= 118 IOPS  (formerly 134)   =>  	IOPS 	= 1/(3.9ms + 2.1ms + .5MB/42.5MBps = 11.8 ms + 1ms CO)  		= 53 IOPS   (formerly 56)   =  	IOPS 	= 1/(3.9ms + 2.1ms + 1MB/42.5MBps = 23.5 ms + 1ms CO)i  		= 33 IOPS   (formerly 34)   ;  	IOPS 	= 1/(3.9ms + 2.1ms + 2MB/42.5MBps = 47 ms + 1ms CO) 1  		= 18.5 which rounds to 19 IOPS   (formerly 19)c  e    * > 	A little something on command overhead: >  > M > IOPS Delivery - The theoretical random IOPS that a typical high performance=N > SCSI disk drive can deliver is the sum of an average seek (9 ms.), a latencyJ > period (4 ms.), and a 1 ms. delay that takes into account transfer time,P > command handshaking and decode. Therefore, you have 9+4+1=14 ms. per transfer.O > 1000 ms./14 ms.=71.43 IOPS per disk. This number assumes that the transfer isCP > 8K or less. Larger I/O's will result in a reduced IOPS number, smaller ones doP > not go faster since 95% of the overhead is either mechanical or command time.  > 
 > 	64 KB I/O:S > 6 > 	IOPS 	= 1/(3.9ms + 2.1ms + 64KB/42.5MBps = 1.47 ms) > 		= 134 IOPS > D > 	Drop off is noticeable but much better IOPS with larger transfers > 	on newer disks. >  > 	By the way, your claim: R > L >> nowhere nearly inversely proportional to the I/O size until the I/O sizes >> reach megabytes in size.I >  > 6 > 	IOPS 	= 1/(3.9ms + 2.1ms + .5MB/42.5MBps = 11.8 ms)
 > 		= 56 IOPSn > 5 > 	IOPS 	= 1/(3.9ms + 2.1ms + 1MB/42.5MBps = 23.5 ms)P
 > 		= 34 IOPSI > 3 > 	IOPS 	= 1/(3.9ms + 2.1ms + 2MB/42.5MBps = 47 ms)I
 > 		= 19 IOPS= > & > 	Close to inverse from 1 MB to 2 MB. > @ > 	And Bill, it isn't always about being clueless or overlookingI > 	or not paying attention or whatever.  Sometimes it is about missing a TH > 	piece or in this case the formula to calculate all this.  Between us A > 	we could have probably looked at that Oracle chart and said:  NC > 	"oh, those must be 7200 RPM SCSI drives" and also concluded theyNE > 	couldn't get more raw or block I/O out of them if the Queen MotherTI > 	wanted it.   I couldn't have done that yesterday, I can do that today.R > 	 > 				RobR >    ------------------------------    Date: 28 Jul 2000 03:24:48 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: I/Os per second, how do you increase them?I+ Message-ID: <a5kk5vXNcKbZ@eisner.decus.org>P  X In article <bqPAaO6oP1cC@eisner.decus.org>, young_r@eisner.decus.org (Rob Young) writes:Z > In article <J+LLM+sLeNVo@eisner.decus.org>, young_r@eisner.decus.org (Rob Young) writes: >  >> T< >> 	Looking at the Seagate 15K disk due out later this year: >> PU >> http://www.seagate.com/cda/products/discsales/enterprise/family/0,1130,246,00.htmlj >> /4 >> 	IOPS 	= 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms); >> 		= 160 IOPS  (I read 140, so I am off here but there isH+ >> 				command overhead so not a one to onej >> 				correspondence) >> i >   > 	One last swipe for Deja sake: > ) > 	Adding 1 ms for command overhead = CO:g > = >  	IOPS 	= 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms + 1ms CO)0@ >  		= 142 IOPS  (formerly 160, there is the 140 we read about). >     : 	I can get away with adding 1 ms to the above, but not the@ 	below.. as CO takes into account decode, handshake and transfer< 	time... so 64KB transfers have IOPS between 118 and 134 and? 	probably more towards 134 than 118 and as we see CO has littlet% 	effect as transfer size gets larger.e     >  	64 KB I/O: >  p@ >  	IOPS 	= 1/(3.9ms + 2.1ms + 64KB/42.5MBps = 1.47 ms + 1ms CO) >  		= 118 IOPS  (formerly 134)- >     @ >  	IOPS 	= 1/(3.9ms + 2.1ms + .5MB/42.5MBps = 11.8 ms + 1ms CO) >  		= 53 IOPS   (formerly 56) >  e? >  	IOPS 	= 1/(3.9ms + 2.1ms + 1MB/42.5MBps = 23.5 ms + 1ms CO)c >  		= 33 IOPS   (formerly 34) >  p= >  	IOPS 	= 1/(3.9ms + 2.1ms + 2MB/42.5MBps = 47 ms + 1ms CO) 3 >  		= 18.5 which rounds to 19 IOPS   (formerly 19)  >      				Robo   ------------------------------  % Date: Fri, 28 Jul 2000 05:22:35 -0400-' From: "Bill Todd" <billtodd@foo.mv.com>h7 Subject: Re: I/Os per second, how do you increase them?l( Message-ID: <8lrj9m$f3t$1@pyrite.mv.net>  5 Rob Young <young_r@eisner.decus.org> wrote in message % news:J+LLM+sLeNVo@eisner.decus.org...aL > In article <8lr1bp$1n0$1@pyrite.mv.net>, "Bill Todd" <billtodd@foo.mv.com> writes:  > > 9 > > Rob Young <young_r@eisner.decus.org> wrote in messagen) > > news:kwzvSbaUAGhW@eisner.decus.org....   ...y  B > >> That chart shows 35 ms for 127 KB transfer.  A 64 KB transfer@ > >> takes 16 ms as the Dell reference points out and the OracleG > >> reference shows.  My first post contained an error as you spotted.AD > >> Divide by two error.  127 blocks makes for a 64 KB I/O , double: > >> the number of I/Os per second but halve the I/O size. > >KI > > Nope - you still have no clue, and won't until you figure out how theOF > > numbers relate to each other.  For a start, the number of I/Os per	 second is_G > > nowhere nearly inversely proportional to the I/O size until the I/OT sizesl > > reach megabytes in size. > >! >p > Uhh.. not quite.  D Yes, quite - for contemporary disks, which is what I've kept stating throughout this discussion.W  )   Doubling the I/O size according to that.$ > chart doubles the transfer time...  L Of course doubling the I/O size doubles the transfer time.  But the transferG time doesn't even *equal* the average-seek-plus-half-rotation time on aSI modern disk until the transfer size exceeds 250 KB (to take a 7200 rpm, 7 K ms. average seek, 4+ ms. half-rotation, median-25 MByte/second disk - which H is reasonably representative of the mid-range of current offerings), and@ doesn't get to 90% of the total time (at which point the inverseJ relationship between transfer size and IOPS is reasonably close to linear)% until the transfer size is over 2 MB.l    The chart reflectsR< > older disks and does make it lean more towards an inverse. >EL > > You can also ignore the Dell numbers (which were specifically offered as anI > > *example*, but happen to be close to an order of magnitude lower than  theRK > > transfer rates of current disks):  a 64 KB transfer takes between 2 andN 3C# > > ms. on modern disks, not 16 ms.K > >I >S# > The formula to calculate IOPS is:U >TL >         Maximum IOPS per disk drive = 1/(seek + rotation + transfer times) >M3 > Plugging in some values for an older disk (SCSI):A > I >                                 = 1/(8 ms + 0.5/7200 RPM + 16KB/6 MBps)mE >                                 = 1/(8 + 4.2 + 2.7) = 1/(14.9 msec)P+ >                                 = 67 IOPS   I At last something I can agree with!  (Perhaps Tim Shoppa will notice thatsH it's the same definition I gave him, and which you appear to corroborate+ later below from an unidentified source...)R   >T> > 67 IOPS for 16 KB I/O size on an older disk.  The numbers in? > the Oracle chart look like older disks.  I suspect John NixonP: > and many of us have RZ29 disks, he is working with HSJs, > 7200 RPM SCSI disks. >P: > Looking at the Seagate 15K disk due out later this year: >M >IL http://www.seagate.com/cda/products/discsales/enterprise/family/0,1130,246,0 0.html > 1 > IOPS = 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms)S8 > = 160 IOPS  (I read 140, so I am off here but there is& > command overhead so not a one to one > correspondence)g  I Actually, a roughly 1 ms. command overhead (which I mentioned previously,GI and which also seems to be corroborated by your unidentified quote below)O( just about accounts for that difference.   >U) > A little something on command overhead:M >A >MA > IOPS Delivery - The theoretical random IOPS that a typical highT performanceIF > SCSI disk drive can deliver is the sum of an average seek (9 ms.), a latencyRJ > period (4 ms.), and a 1 ms. delay that takes into account transfer time,F > command handshaking and decode. Therefore, you have 9+4+1=14 ms. per	 transfer.PL > 1000 ms./14 ms.=71.43 IOPS per disk. This number assumes that the transfer isH > 8K or less. Larger I/O's will result in a reduced IOPS number, smaller ones do-I > not go faster since 95% of the overhead is either mechanical or commandn time.s >e > 64 KB I/O: >r4 > IOPS = 1/(3.9ms + 2.1ms + 64KB/42.5MBps = 1.47 ms) > = 134 IOPS >eC > Drop off is noticeable but much better IOPS with larger transfers> > on newer disks.d >s > By the way, your claim:  >tG > > nowhere nearly inversely proportional to the I/O size until the I/O  sizes  > > reach megabytes in size. >. > 4 > IOPS = 1/(3.9ms + 2.1ms + .5MB/42.5MBps = 11.8 ms) > = 56 IOPSa > 3 > IOPS = 1/(3.9ms + 2.1ms + 1MB/42.5MBps = 23.5 ms)r > = 34 IOPS4 >/1 > IOPS = 1/(3.9ms + 2.1ms + 2MB/42.5MBps = 47 ms)  > = 19 IOPSt >1% > Close to inverse from 1 MB to 2 MB.-  J 'Close' is of course in the eye of the beholder.  And technically anythingL exceeding 1 MB is 'megabytes'.  I made the statement as an emphatic responseK to "127 blocks makes for a 64 KB I/O , double the number of I/Os per secondf but halve the I/O size":  1 1/(3.9 ms. + 2.1 ms. + 127KB/42.5MBps) = 111 IOPS-  0 1/(3.9 ms. + 2.1 ms. + 64KB/42.5MBps) = 133 IOPS  I You don't need numbers from a super-high-end drive:  current 7200 rpm IDEfK drives produce over 20 MBytes/second from the innermost tracks and close to>+ 40 MBytes/second from the outermost tracks:s  , 1/(7 ms. + 4.2 ms. + 127KB/30MBps) = 65 IOPS  + 1/(7 ms. + 4.2 ms. + 64KB/30MBps) = 75 IOPS4  G "double the number of I/Os per second but halve the I/O size"?  I don'ttJ think so:  it's nowhere near linear.  And even less so if you add in the 1L ms. command overhead (though since that can often be hidden during execution5 of the previous operation I'll not press that point).e   > ? > And Bill, it isn't always about being clueless or overlooking G > or not paying attention or whatever.  Sometimes it is about missing ap: > piece or in this case the formula to calculate all this.  L So if you don't know it, ask, rather than assume it's irrelevant when you'veK been told otherwise.  Arguing about numbers you *know* you don't understand2 is unlikely to be productive.      Between us> > we could have probably looked at that Oracle chart and said:B > "oh, those must be 7200 RPM SCSI drives" and also concluded theyD > couldn't get more raw or block I/O out of them if the Queen MotherH > wanted it.   I couldn't have done that yesterday, I can do that today.  L It's more significant that they're relatively old drives (with significantlyK lower bit-densities) than 7200 rpm drives:  current 7200 rpm drives deliver ( around 30 MBytes/second, as noted above.  J But what you *really* need to look at is the *pattern* of times - which is; what caused me to make my initial comments on this subject.:  K Start with the sequential read results in Table 15 - 7.  For small (8 KB ornK less) reads, the Unix File System column suggests transfer rates of about 8oF MBytes/second (plus a fraction of a millisecond constant).  For largerG reads, the transfer rate for the portion beyond the first 8 KB drops toqI between 3 and 4 MBytes/second.  Then when you go from 128 KB to 256 KB in>L the last step, the last 128 KB of the 256 KB read bounces back up to about 5F MBytes/second - which would be my best guess at something close to theK *real* streaming performance of the drive, since there's too much potentialsK variance in the very small transfer times to be sure of the 8 MBytes/secondoH rate.  (The article copyright is 1997, and the disks might not have been? high-end, so something like 5 - 6 MBytes/second is believable.)s  H The 'block device' and 'raw device' columns are considerably more linearJ (though the 'raw device' has a noticeable step when you exceed 4 KB) - andI *always* noticeably *slower* than the Unix File System comparable figure, 3 though still indicative of 'streaming' operation...C  K The sequential write results in Table 15 - 8 shed more light on the matter. B The 'raw device' column is roughly linear in write size (with a 3+F MBytes/second apparent transfer rate - though this drops off at 64+ KBH transfer sizes), plus an 11 ms. constant that rather strongly suggests aK 5400 rpm disk that misses a rev on each write (rather than streaming them).oJ (And a 5400 rpm disk with around 100 blocks/track would generate streamingG data rates of around 6 MBytes/second, which is consistent both with therA analysis above and with average density increases over 3+ years.)a  I The 'block device' column is almost identical up to 8 KB, then dives in ahK manner that strongly suggests that within each individual write request thedI disk is missing a rev (actually, a bit more than a rev - but the need forlH interleaving disappeared with the advent of fast on-disk controllers, so@ that will have to remain a mystery) after every 8 KB transfered.  G The 'Unix File System' column is almost identical to the 'block device'mI column save for an additional constant 7 ms. or so (though above 64 KB it  gains a bit of ground back).  J Move to the random read table (15 - 9) and things get even more squirrely:B the 'block device' column has a major step after 4 KB plus anotherJ noticeable change (to quasi-linear increases) after 32 KB, while the 'UnixG File System' column has a major step after 8 KB plus another noticeablefK change (to quasi-linear increases) after 64 KB.  And while the 'raw device'lL column is reassuringly linear (plus a 12 ms. constant), it reflects transfer, rates of only between 2 and 3 MBytes/second.  I And the random write table (15 - 10) would be totally inexplicable if youaK didn't realize that both the 'block device' and the 'Unix File System' must C be using 8 KB access units, so that all smaller writes must perform4L read/modify/write sequences.  And, as with sequential writes, each 8 KB of aL large single random write request seems to take an extra 13 - 15 ms. or so -K same mystery as before - and clearly a peculiarity of the system under test I rather than anything intrinsic to disk performance on a system capable of( exploiting it.  D All of which is why I made my very first comment in this discussion:  F "That chart bears little relation to both current disk performance andI non-Unix environments.  In other words, it's useless in this discussion."   , which is just as correct now as it was then.  L I will admit that I can't be certain that the cause of the lousy performanceL (compared to the base 5 MBytes/second that can be seen in the best cases) isL 'brain-damaged' old-Unix-style allocation rather than to some other quirk ofK the system (and from the above analysis there appear to be several).  AsiderB from that statement (which I referred to again later - mea culpa),K everything which I have said (and which you have persisted in arguing with) I remains 100% valid - and if I've been less than tactful in presenting it,sG it's because I don't particularly appreciate being contradicted about a	0 subject I'm as qualified to discuss as this one.   - bill   >i > Rob  >a   ------------------------------  % Date: Fri, 28 Jul 2000 05:26:37 -0400r' From: "Bill Todd" <billtodd@foo.mv.com>e7 Subject: Re: I/Os per second, how do you increase them?t( Message-ID: <8lrjh7$fbd$1@pyrite.mv.net>  5 Rob Young <young_r@eisner.decus.org> wrote in message0% news:a5kk5vXNcKbZ@eisner.decus.org...eK > In article <bqPAaO6oP1cC@eisner.decus.org>, young_r@eisner.decus.org (Robn Young) writes:H > > In article <J+LLM+sLeNVo@eisner.decus.org>, young_r@eisner.decus.org (Rob Young) writes:- > >  > >>= > >> Looking at the Seagate 15K disk due out later this year:  > >> > >>L http://www.seagate.com/cda/products/discsales/enterprise/family/0,1130,246,0 0.html > >>4 > >> IOPS = 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms); > >> = 160 IOPS  (I read 140, so I am off here but there isg) > >> command overhead so not a one to one  > >> correspondence) > >> > >c! > > One last swipe for Deja sake:e > >l* > > Adding 1 ms for command overhead = CO: > >m= > >  IOPS = 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms + 1ms CO) @ > >  = 142 IOPS  (formerly 160, there is the 140 we read about). > >  >	 >r; > I can get away with adding 1 ms to the above, but not the A > below.. as CO takes into account decode, handshake and transfer4	 > time...   K No, it's equally applicable in both places.  The inclusion of transfer time.H in the article's comment was for 8 KB and smaller transfers (i.e., thoseK where the transfer time was down in the noise level):  the command overhead.L and hand-shaking are most of the 1 ms., and since transferring 64 KB at 42.56 MBytes/second takes 1.5 ms. it's clearly not included.   - bill  4  so 64KB transfers have IOPS between 118 and 134 and@ > probably more towards 134 than 118 and as we see CO has little& > effect as transfer size gets larger. >  >c > >  64 KB I/O:  > > @ > >  IOPS = 1/(3.9ms + 2.1ms + 64KB/42.5MBps = 1.47 ms + 1ms CO) > >  = 118 IOPS  (formerly 134)  > >e >e@ > >  IOPS = 1/(3.9ms + 2.1ms + .5MB/42.5MBps = 11.8 ms + 1ms CO) > >  = 53 IOPS   (formerly 56) > > ? > >  IOPS = 1/(3.9ms + 2.1ms + 1MB/42.5MBps = 23.5 ms + 1ms CO)r > >  = 33 IOPS   (formerly 34) > >.= > >  IOPS = 1/(3.9ms + 2.1ms + 2MB/42.5MBps = 47 ms + 1ms CO)a3 > >  = 18.5 which rounds to 19 IOPS   (formerly 19)e > >i >  > Rob  >S   ------------------------------  # Date: Fri, 28 Jul 2000 12:16:20 GMTt/ From: "John Nixon" <jorlnixon@worldnet.att.net>O7 Subject: Re: I/Os per second, how do you increase them?7E Message-ID: <oQeg5.1222$ZL5.63881@bgtnsc07-news.ops.worldnet.att.net>S  @ This whole conversation reminds me of an old saying that I like,  >     "If two people agree all the time, one of them is useless"H     "If two people DISagree all the time, then both of them are useless"  2 "Bill Todd" <billtodd@foo.mv.com> wrote in message" news:8lr4iu$46r$1@pyrite.mv.net... > 7 > Rob Young <young_r@eisner.decus.org> wrote in message=' > news:xBAe+$eEhdNO@eisner.decus.org...I8 > > In article <8lq6sa$6an$1@pyrite.mv.net>, "Bill Todd" <billtodd@foo.mv.com> 	 > writes:s > > >b; > > > Rob Young <young_r@eisner.decus.org> wrote in messageh+ > > > news:hNOk6DhVGuJR@eisner.decus.org... ; > > >> In article <8lq3om$3v6$1@pyrite.mv.net>, "Bill Todd"t > <billtodd@foo.mv.com>e
 > > > writes:d > > >>J > > >> >> > whatsoever in presenting them as established fact.  After all, > just	 > > > howN > > >> > do L > > >> >> > you expect to get around 100 I/O operations/second (numbers that > > > you've> > > >> >>                            ^^^^^^^^^^^^^^^^^^^^^^^^^G > > >> >> Simple.  Use smaller I/Os and most numbers you see quoted arem > > >> >> based on 2-4K I/Os.  > > >> >K > > >> > Rob, Rob, Rob:  will you *ever* pay attention.  Please re-read thec > next# > > >> > paragraph you quote below.o > > >> > > > >> > > >>A > > >> Doesn't have to do with paying attention.  Maybe I have itN' > > >> right and you are wrong.  Maybe?s > > >oH > > > Not a chance.  But if you really are paying attention, you need to starts > > > working on comprehension.t > > >d > >d2 > > Gee, haven't had so much fun in quite a while. > >/9 > > In the larger context of things you wrote this and my	 > > reply is interspersed: > > H > > > 2 - 3 ms., for a total of 10 - 14 ms.  Add a millisecond or so for	 > gettingnD > > > the command through the SCSI adapter and at most a few hundred > microsecondsK > > > for OS processing and the result is 11 - 15 ms. - which is consistentO > withD > > > enough real-world performance results I've seen that I have no > hesitationI > > > whatsoever in presenting them as established fact.  After all, justo howe > doF > > > you expect to get around 100 I/O operations/second (numbers that you've8 > >                            ^^^^^^^^^^^^^^^^^^^^^^^^^I > >         Simple.  Use smaller I/Os and most numbers you see quoted are  > >         based on 2-4K I/Os.+ > >/J > > > quoted elsewhere) out of a disk if it takes 1/30th of a second to do > each
 > > > one? > >+: > > At which point, you try to deflect things but don't do	 > > well.e > F > The trouble is, when you don't understand something, you assume it'sH > irrelevant.  And quantitative analysis of this situation appears to be& > completely beyond your capabilities. >s > >9= > > How do you get 100 IOPS out of a disk?  Use smaller I/Os.- >-J > The point, Rob, was that if your contention was valid (which it was not)J > that 64 KB I/O operations took 35 ms., then you could reduce them to 512L > bytes and it would do no more than improve your IOPS rate from about 29 toK > about 32 - nowhere near 100.  (And now that you've changed your number to1K > 127 KB for the request size, I'll point out that if a 127 KB request took  35L > ms., then reducing it to 512 bytes would improve your IOPS rate from about+ > 29 to about 35 - still nowhere near 100.)  >wI > That was not deflection, it was an attempt to point out just how absurdl yourG > contention was by showing that it was inconsistent with other numbersa (80 - G > 100+ IOPS) you yourself have been quoting lately (and *those* numbers  aren't > far removed from reality). >OK > I'm not sure why you seem to believe that disk performance is a matter of,L > opinion (and hence endlessly arguable) rather than measurement.  But untilC > you can understand and apply the measured performance of disks tovK > relationships such as that of IOPS vs. transfer size discussed above, youe" > have nothing to contribute here. >. > - bill >e >   If; > > your I/O size is 127 KB , you won't get 100 IOPS out oft > > a 7200 RPM disk, nor 10K.g > >I > > Robc > >n >o >a   ------------------------------    Date: 28 Jul 2000 09:47:48 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: I/Os per second, how do you increase them?a+ Message-ID: <XaOupxnK$n7i@eisner.decus.org>   w In article <oQeg5.1222$ZL5.63881@bgtnsc07-news.ops.worldnet.att.net>, "John Nixon" <jorlnixon@worldnet.att.net> writes: B > This whole conversation reminds me of an old saying that I like, > @ >     "If two people agree all the time, one of them is useless"J >     "If two people DISagree all the time, then both of them are useless" >   : 	But we don't , no more than Andrew and I disagree all the/ 	time, Bill and I agree quite often actually.  t  H 	Maybe it is a bit much but go back and read some of the latter posts.  G 	I know now why the "Tuning I/O" Oracle document's bar graphs look the  H 	way they do.   They are using old disks in their comparison.  Shame on B 	them for not noting that (may have missed that)... and quite the D 	bother they aren't using disks that have been around for at *least* 	the last 3 years or so.  : 	How can I make the statement in the first line?  Oh.. you: 	may be surprised and I anticipate you would agree with me7 	that is a true statement.  Can't say much more now ;-)    				Rob    ------------------------------    Date: 28 Jul 2000 10:42:12 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: I/Os per second, how do you increase them?a+ Message-ID: <8Pk0y5uRiiJq@eisner.decus.org>I  R In article <8lrjh7$fbd$1@pyrite.mv.net>, "Bill Todd" <billtodd@foo.mv.com> writes: > 7 > Rob Young <young_r@eisner.decus.org> wrote in message ' > news:a5kk5vXNcKbZ@eisner.decus.org...iL >> In article <bqPAaO6oP1cC@eisner.decus.org>, young_r@eisner.decus.org (Rob > Young) writes:I >> > In article <J+LLM+sLeNVo@eisner.decus.org>, young_r@eisner.decus.orga > (Rob Young) writes:  >> > >> >>a> >> >> Looking at the Seagate 15K disk due out later this year: >> >>  >> >>ON > http://www.seagate.com/cda/products/discsales/enterprise/family/0,1130,246,0 > 0.html >> >>h5 >> >> IOPS = 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms)s< >> >> = 160 IOPS  (I read 140, so I am off here but there is* >> >> command overhead so not a one to one >> >> correspondence)  >> >>  >> >" >> > One last swipe for Deja sake: >> >+ >> > Adding 1 ms for command overhead = CO:i >> >> >> >  IOPS = 1/(3.9 + 2.1ms + 4KB/42.5MBps = < .1 ms + 1ms CO)A >> >  = 142 IOPS  (formerly 160, there is the 140 we read about).  >> > >> >>< >> I can get away with adding 1 ms to the above, but not theB >> below.. as CO takes into account decode, handshake and transfer
 >> time... > M > No, it's equally applicable in both places.  The inclusion of transfer time1J > in the article's comment was for 8 KB and smaller transfers (i.e., thoseM > where the transfer time was down in the noise level):  the command overheadfN > and hand-shaking are most of the 1 ms., and since transferring 64 KB at 42.58 > MBytes/second takes 1.5 ms. it's clearly not included. >  > - bill > 6 >  so 64KB transfers have IOPS between 118 and 134 andA >> probably more towards 134 than 118 and as we see CO has littler' >> effect as transfer size gets larger.y >>    ; 	Nice.  Learning can be a very painful experience and whiley9 	this has been quite painful I actually have already usedM; 	it to understand another issue.  Similarly, Andrew has had 9 	me jump through hoops and I have had to eat crow on morem7 	than one occasion with him too.  (Hint:  if you dredgel7 	up tpmC stuff and try to manipulate it , you better bee= 	spot on or Andy will spin you 'round and 'round.)  Likewise,n= 	if you start touching disk and I/O and filesystem ... betterp 	be spot on there too :-).   				RobI   ------------------------------  % Date: Fri, 28 Jul 2000 16:14:53 +01003- From: Tim Llewellyn <tim.llewellyn@bbc.co.uk> 7 Subject: Re: I/Os per second, how do you increase them? ) Message-ID: <3981A36D.FB5005D6@bbc.co.uk>i   Rob Young wrote:  A >          At which point, you try to deflect things but don't dos >         well.0 >yG >         How do you get 100 IOPS out of a disk?  Use smaller I/Os.  IfpA >         your I/O size is 127 KB , you won't get 100 IOPS out ofu# >         a 7200 RPM disk, nor 10K.  >v  A So, this has got me intruiged, why are you guys, and VMS monitor, > so hooked on IO's per second rather than megabytes per second?> Is it a metric in common use in the spin^H^H^H^Htrade press or= what? Why? Sure, I can see that the throughput depends on thenB IO size as Rob points out. In fact this is so obvious it should be taken as given..  < Presumably, the original poster has a known workload and I/O= packet size, and wants to know how to crank more I/O from his  system.a     --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.ukr  A I speak for myself only and my views in no way represent those of0 MedAS or the BBC.a   ------------------------------    Date: 28 Jul 2000 12:40:03 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: I/Os per second, how do you increase them?t+ Message-ID: <fHvTnK0kmmTb@eisner.decus.org>l  Y In article <3981A36D.FB5005D6@bbc.co.uk>, Tim Llewellyn <tim.llewellyn@bbc.co.uk> writes:i  C > So, this has got me intruiged, why are you guys, and VMS monitor,*@ > so hooked on IO's per second rather than megabytes per second?   	They are related.   				Robm   ------------------------------  % Date: Fri, 28 Jul 2000 17:16:30 +01005- From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>i7 Subject: Re: I/Os per second, how do you increase them? ) Message-ID: <3981B1DE.D8779842@bbc.co.uk>c   Rob Young wrote:  [ > In article <3981A36D.FB5005D6@bbc.co.uk>, Tim Llewellyn <tim.llewellyn@bbc.co.uk> writes:  >BE > > So, this has got me intruiged, why are you guys, and VMS monitor,,B > > so hooked on IO's per second rather than megabytes per second? >s >         They are related.t >e  N Sure, I know that, I knew that before I read this thread. The thing is, to getR Mbyte/sec you have to know how large your IO's are, unless I am missing something.S What I am saying is, why does Monitor process/topdio for example not have an optionv& to report Mb/s rather than just I/O's.  L Sorry to jump into this big-boy thread, but I have often wondered about thisI in my years with VMS. I have written test programs to study throughput inp	 the past.x   --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.uki  A I speak for myself only and my views in no way represent those ofm MedAS or the BBC.    ------------------------------  % Date: Fri, 28 Jul 2000 12:27:13 -0400s' From: "Bill Todd" <billtodd@foo.mv.com>u7 Subject: Re: I/Os per second, how do you increase them?0( Message-ID: <8lsc5r$a5c$1@pyrite.mv.net>  5 Rob Young <young_r@eisner.decus.org> wrote in message % news:fHvTnK0kmmTb@eisner.decus.org...s9 > In article <3981A36D.FB5005D6@bbc.co.uk>, Tim Llewellynn! <tim.llewellyn@bbc.co.uk> writes:o >aE > > So, this has got me intruiged, why are you guys, and VMS monitor,'B > > so hooked on IO's per second rather than megabytes per second? >t > They are related.w  L And they're also interesting individually (rough sizes below apply to modern. disks - reduce them somewhat for older disks).  G If your access is primarily characterized by 'streaming' data, then theaG disk's sustainable transfer rate off its platters (MB/second) primarilysI governs your performance (unless something like the bus creates a tighteraL bottleneck).  Even if multiple processes interleave their streaming, as longJ as their individual requests are relatively large (say, a half-megabyte or7 larger) transfer times will dominate positioning times.r  K If your access is primarily characterized by smallish (say, 64 KB or below)wK random requests, then positioning time will dominate transfer time and IOPSn4 is a number of more significance than transfer rate.  G If your access is more varied, then it's the responsibility of the file2L system to do the best it can (e.g., notice the pattern in a series of small,G sequential requests even if they're interleaved with other activity andeK pre-fetch in larger chunks to avoid the frequent positioning overheads, andeK cache random requests for a while in case they're not really random and the4F data is requested again).  Which doesn't make the disk characteristics7 irrelevant but attempts to take best advantage of them.n   - bill   >m > Robo >    ------------------------------  % Date: Fri, 28 Jul 2000 06:59:38 +0200   From: Paul Sture <paul@sture.ch>2 Subject: Re: Interbase 6.0 released as Open Source+ Message-ID: <VA.0000008d.19201e65@sture.ch>f  c In article <009EDB5D.CA2858AA@SSRL04.SLAC.STANFORD.EDU>, Alan Winston - SSRL Admin Cmptg Mgr wrote:aN > From: winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") > Newsgroups: comp.os.vmsn4 > Subject: Re: Interbase 6.0 released as Open Source% > Date: Thu, 27 Jul 2000 21:16:59 GMTm >    > >aF > I think this had a lot to do with Starkey (the original architect ofF > Datatrieve, responsible for wombats and yacht) having designed DSRI. > h AAARRRRGGGGHHHH. Please DROP YACHTS! I got very sick of customers comparing that few of the examples in h the books worked. KIDS and other related stuff were far too US orientated for an international audience f too. Quite apart from taking something like 2 seconds of CPU on a 780 to get to the prompt (read Stan g Rabowitz's article on that) - yes I halved that CPU by installing various of the many images concerned,nY but when I saw a job ad a year or so later for DTR work, I simply decided I was too young0 for a heart attack.   Z Whilst good for certain stuff, it just didn't cut the mustard, for myself and many others.  ? IMHO, the only good thing to come out of it was Stan's article.e ___u
 Paul Sture Switzerland    ------------------------------  # Date: Fri, 28 Jul 2000 07:46:34 GMT.L From: winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr")2 Subject: Re: Interbase 6.0 released as Open Source8 Message-ID: <009EDBB5.BDE375E0@SSRL04.SLAC.STANFORD.EDU>  N In article <VA.0000008d.19201e65@sture.ch>, Paul Sture <paul@sture.ch> writes:d >In article <009EDB5D.CA2858AA@SSRL04.SLAC.STANFORD.EDU>, Alan Winston - SSRL Admin Cmptg Mgr wrote:O >> From: winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr")  >> Newsgroups: comp.os.vms5 >> Subject: Re: Interbase 6.0 released as Open Source & >> Date: Thu, 27 Jul 2000 21:16:59 GMT >> 1 >  >> >G >> I think this had a lot to do with Starkey (the original architect of G >> Datatrieve, responsible for wombats and yacht) having designed DSRI.d >> ti >AAARRRRGGGGHHHH. Please DROP YACHTS! I got very sick of customers comparing that few of the examples in ai >the books worked. KIDS and other related stuff were far too US orientated for an international audience ug >too. Quite apart from taking something like 2 seconds of CPU on a 780 to get to the prompt (read Stan hh >Rabowitz's article on that) - yes I halved that CPU by installing various of the many images concerned,Z >but when I saw a job ad a year or so later for DTR work, I simply decided I was too young >for a heart attack. >B  N >Whilst good for certain stuff, it just didn't cut the mustard, for myself and many others.   >a@ >IMHO, the only good thing to come out of it was Stan's article.    O I'm actually using Datatrieve very happily now, mostly in batch mode to produce N reports and web pages for automatic distribution.  Runs well, easy to code in,L and on an AS2100 5/300 it isn't too bad to run for interactive use.  On the O VAX 11/750 I learned it on, I spent a lot of time figuring out how to optimize.)   -- Alan   O ===============================================================================s0  Alan Winston --- WINSTON@SSRL.SLAC.STANFORD.EDUM  Disclaimer: I speak only for myself, not SLAC or SSRL   Phone:  650/926-3056mM  Physical mail to: SSRL -- SLAC BIN 69, PO BOX 4349, STANFORD, CA  94309-0210 O ===============================================================================n   ------------------------------  # Date: Fri, 28 Jul 2000 15:58:43 GMTI From: jordan@my-deja.com2 Subject: Re: Interbase 6.0 released as Open Source) Message-ID: <8lsajf$liu$1@nnrp1.deja.com>b  4 In article <ae3g5.81$mO3.14794@typhoon.aracnet.com>,5   "Zane H. Healy" <healyzh@shell1.aracnet.com> wrote:.2 > Jordan Henderson <jordan@lisa.gemair.com> wrote:E > > I'd sure love to see Hobbyists using something like this too.  In<E > > the Linux world, Hobbyists are called developers and they do somet > > great stuff. >o> > I'd have a go at it, except for the slight matter of time...  ? Yes, I also have this problem.  I didn't mean to pick on anyone@< as complaining about a situation that they aren't willing to rectify.  D >                                                              I got what IF > consider to be a very cool little program ready to start testing and then IG > ran out of time about 2 months ago, just as I was ready to run it for  the F > first time.  I've not touched it since :^(  The only thing I'm using VMS formC > at the moment is a notepad. :^(  Real life has its advantages and  > disadvantages :^)  >lF > Note, while I'd be willing to have a go at it if I had time, I doubt I'd beD > successful.  My C is so rusty it isn't funny!  So it sure would be nice is : > some skilled individual with extra time gave it a go :^)  @ Well, I probably won't be porting Interbase 6 to OpenVMS either.   Hopefully, someone will...   > 	 > 			Zane  >c   -Jordan Hendersond jordan@greenapple.comt    & Sent via Deja.com http://www.deja.com/ Before you buy.o   ------------------------------  % Date: Fri, 28 Jul 2000 08:35:04 +0100 5 From: "Chris Casey" <Chris.Casey@HendersonNOSPAM.com>e8 Subject: Re: is there a database available for VMS hobby) Message-ID: <8lrd3g$6n8$1@soap.pipex.net>i  ' Take a look at Cache from Intersystems. G This is one of the fastest and most scaleable databases around for VMS.oK It's predecessor DSM used to produce benchmark figures that scared the hellv out of Oracle and Rdb. (www.intersys.com)  " Zane H. Healy wrote in message ..., >Glenn C. Everhart <Everhart@gce.com> wrote:3 >> There are a few dbms systems that are available. ; >> DBAG is a dBase III sorta-clone dbms for example. RIM isr= >> an adequate single user dbms (also available interfaced ton( >> AnalytiCalc spreadsheet as Analyrim). >uL >Anyone have any real pointers to these?  I'm assuming RIM isn't part of theJ >Analyrim package.  It looks like the only place they might be are on some ofH >the DECUS tapes, which brings up another question, is there any kind of list# >of what is on the DECUS VMS tapes?t >i< >> There are a number of simpler packages around also. For a> >> good multi-user dbms, I suspect porting PostgreSQL might be> >> one of the better alternatives. It is related to university= >> Ingres; there was (maybe is sorta: CA got it) a commercialt/ >> Ingres that at one time at least ran on VMS.w >dJ >I actually took a look at PostgreSQL or whatever it's called these days aF >few months ago...  Only I can't seem to remember what I found...  HowJ >irritating!  I know the source is still in my user directory.  I also wasJ >thinking mySQL might be a candidate.  Overall though I think it sounds as ifJ >Interbase might be the best candidate since it has at least run on VMS at >one point in its life.u >h > Zane >    ------------------------------  % Date: Fri, 28 Jul 2000 10:29:54 -0700r7 From: "Arthur E. Ragosta" <ragosta@merlin.arc.nasa.gov>a8 Subject: Re: is there a database available for VMS hobby3 Message-ID: <3981C312.AADD1159@merlin.arc.nasa.gov>a   "Glenn C. Everhart" wrote:  2 > There are a few dbms systems that are available.: > DBAG is a dBase III sorta-clone dbms for example. RIM is< > an adequate single user dbms (also available interfaced to' > AnalytiCalc spreadsheet as Analyrim).d >o; > There are a number of simpler packages around also. For ae= > good multi-user dbms, I suspect porting PostgreSQL might beo= > one of the better alternatives. It is related to universityo< > Ingres; there was (maybe is sorta: CA got it) a commercial. > Ingres that at one time at least ran on VMS.   Glenn,  ; I've lost my copy of RIM source.  Do you have a copy on CD,g@ 8mm, or 4mm?  If so, I'd consider looking at it, cleaning it up,> and posting it on my web site.  May take a while... after all,9 Kronos version III has been waiting for 2 years now.  ;-)    ------------------------------  % Date: Fri, 28 Jul 2000 09:27:51 -0700l! From: Shane.F.Smith@Healthnet.com ! Subject: Re: Memory & CPU SupporthC Message-ID: <OFE684F4B5.B65BBC48-ON8825692A.005A5302@HEALTHNET.COM>   F Hey, I could keep 32 busy right now with code I've already written. OfC course,  to hit the 32 processor mark, management's projected usagee* estimates would have to be right...... ;-)   Shanea          > David A Froble <davef@tsoft-inc.com> on 07/27/2000 12:30:56 PM   To:   Info-VAX@Mvb.Saic.Come cc:   " Subject:  Re: Memory & CPU Support     "Bart Z. Lederman" wrote:b >e: > Actually coming up with an application that will keep 32@ > processors busy at the same time is another matter altogether.@ > I would be very interested to hear from anyone who thinks they? > have such an application.  We have seen applications that user= > up to 14 processors on existing platforms, but I'm not suree > they'd scale to 32.-  K Well, to take this to the extreme, I think I could design an application to  keepK 32 processors busy, but, it would be designed to do just that, and probablyc6 wouldn't have much (any) real world applicability. :-)   Dave   --4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Fri, 28 Jul 2000 06:59:36 +0200 $ From: Paul Sture <sture.ch@sture.ch>) Subject: Re: Microvax 3100-10 and 20  $99>+ Message-ID: <VA.0000008c.192016ec@sture.ch>o  C In article <398092D2.9270404E@tsoft-inc.com>, David A Froble wrote: ' > Date: Thu, 27 Jul 2000 15:51:46 -0400., > From: David A Froble <davef@tsoft-inc.com> > Newsgroups: comp.os.vms>+ > Subject: Re: Microvax 3100-10 and 20  $99  >   @ > 11 unneeded posts.  Once again, what's the definition of spam? > X On the subject of spam, can I ask if anyone else here is getting sick of the same tired W posts from ted/joe/henry/tim/whoever, but mainly 3 letter names @zzn.com. It's getting   quite tiresome.> ___n
 Paul Sture Switzerlands   ------------------------------    Date: 28 Jul 2000 07:45:22 -05009 From: Kilgallen@eisner.decus.org.nospam (Larry Kilgallen)y) Subject: Re: Microvax 3100-10 and 20  $99o+ Message-ID: <cA3VHmEEZjC9@eisner.decus.org>w  Z In article <398092D2.9270404E@tsoft-inc.com>, David A Froble <davef@tsoft-inc.com> writes:  8 > So, at the time I'm writing this, the score stands at: > ? > 	1 - announcement of some rather cheap prices for some old HW  > 	1 - defense of original postc( > 	2 - rather neutral posts on the topic  > 	3 - against the original post > 	4 - for the original post  N > Some of the above actually could have been classified a bit differently, butQ > hey, it's my post.  In any case, including my comments, 11 posts to comp.os.vms J > that wouldn't have existed if the first was ignored by those who weren'tM > interested.  As for my stance, I'm sure there are still a few who'd like antQ > inexpensive home system, and maybe even a few who'd consider them for somethings > else.t > @ > 11 unneeded posts.  Once again, what's the definition of spam?  F You ignore the effect of Islandco Wannabees who might see the originalF spam go unchallenged, decide this group does not consider it spam, and start hawking their own wares.  F I will concede that some spammers do not read the groups they spam and( thus do not care about their reputation.   ------------------------------  % Date: Fri, 28 Jul 2000 15:34:21 +0200o= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>n) Subject: Re: Microvax 3100-10 and 20  $99 ) Message-ID: <39818BDD.76417033@gtech.com>?   Larry Kilgallen wrote:\ > In article <398092D2.9270404E@tsoft-inc.com>, David A Froble <davef@tsoft-inc.com> writes:: > > So, at the time I'm writing this, the score stands at: > >,F > >       1 - announcement of some rather cheap prices for some old HW& > >       1 - defense of original post/ > >       2 - rather neutral posts on the topicr' > >       3 - against the original posto# > >       4 - for the original post  > P > > Some of the above actually could have been classified a bit differently, butS > > hey, it's my post.  In any case, including my comments, 11 posts to comp.os.vmsOL > > that wouldn't have existed if the first was ignored by those who weren'tO > > interested.  As for my stance, I'm sure there are still a few who'd like an.S > > inexpensive home system, and maybe even a few who'd consider them for something 	 > > else.e > >tB > > 11 unneeded posts.  Once again, what's the definition of spam? > H > You ignore the effect of Islandco Wannabees who might see the originalH > spam go unchallenged, decide this group does not consider it spam, and  > start hawking their own wares.  G 1)  islandco has been posting here for >1 year and so far they have hadk nott     had any real followers  B 2)  if the followers started posted info about cheap alphas, cheap
 alpha-memory,eG     cheap disks and tape-drives for VMS system, then I would not care -- actually%     I think it would be great for VMS1  H > I will concede that some spammers do not read the groups they spam and* > thus do not care about their reputation.  ! That would have been my point #3.c   Arne   ------------------------------  % Date: Fri, 28 Jul 2000 15:38:06 +0200l= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>t( Subject: Re: Microvax 3100-10 and 20 $99) Message-ID: <39818CBD.BFB1C025@gtech.com>r   Wayne Sewell wrote:rS > In article <8lp4q8$9mf$1@nnrp1.deja.com>, Alan Greig <agreig@my-deja.com> writes:r2 > > In article <fq5e9yXwi8lC@tachxxsoftxxconsult>,D > >   wayne@tachysoft.xxx.293778.killspam.0223 (Wayne Sewell) wrote:F > >> Yes, if you mean posting advertisements to a technical newsgroup,
 > > again ando: > >> again and again and again and again and again and ... > >tI > > I know there are differing views here but we bought four Multias fromVK > > Island for home hobby VMS use at little more than shipping cost and I'diK > > have never known about this particular source other than via occasionall > > postings here. > F > occasional?  Island could teach Compaq about the meaning of NonStop.  C If you read this thread, then you will see that many people do findA! islandco's offerings interesting.-  - My guess is that many people prioritize like:0   1)  good technical infol" 2)  good offers for cheap hardware) 3)  long threads about what Compaq shoulde 4)  OS & system wars   Arne   ------------------------------   Date: 28 Jul 2000 09:37:19 CDT= From: wayne@tachysoft.xxx.293778.killspam.0223 (Wayne Sewell)F( Subject: Re: Microvax 3100-10 and 20 $99. Message-ID: <LH3l7p79bluc@tachxxsoftxxconsult>  [ In article <3980F67A.E097477E@bigpond.com>, David B Sneddon <dbsneddon@bigpond.com> writes:e > Wayne Sewell wrote:c >>   >>  G >> occasional?  Island could teach Compaq about the meaning of NonStop.z >> l > F > Are you suggesting that maybe Island take over the Stealth Marketing > Arm of Compaq? >   L God, no!  Right now, compaq and vms are merely being ignored.  We don't want them to be actively despised.m   --  O ===============================================================================eM Wayne Sewell, Tachyon Software Consulting  (281)812-0738  wayne@tachysoft.xxxr: http://www.tachysoft.xxx/www/tachyon.html and wayne.html  K change .xxx to .com in addresses above, assuming you are not a spambot  :-)cO ===============================================================================tO Otter, on dining with Bluto:"It's perfectly safe if you keep your arms and legsr 			away from his mouth."   ------------------------------  % Date: Fri, 28 Jul 2000 06:59:40 +0200t  From: Paul Sture <paul@sture.ch>& Subject: Re: MS IntelliMouse on Alpha?+ Message-ID: <VA.00000090.19202766@sture.ch>y  J In article <8lpngu$fic$2@mailint03.im.hou.compaq.com>, Hoff Hoffman wrote:4 > From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman) > Newsgroups: comp.os.vmss( > Subject: Re: MS IntelliMouse on Alpha?  > Date: 27 Jul 2000 16:20:46 GMT >  > u > In article <8lpbdj$6ge$2@charm.magnus.acs.ohio-state.edu>, DAVISM@er6.eng.ohio-state.edu (Michael T. Davis) writes:-G > : Can the IntelliMouse (2-button with scroll wheel/button) be used on P > :an Alpha workstation?  Presumably, the wheel functionality would be lost, butP > :could the wheel "button" still be used?  If so, how do the two standard mouseM > :buttons and the wheel button on the IntelliMouse map to the standard threeoK > :buttons on the typical Alpha mouse?  Alternately, could a standard AlphasO > :workstation mouse be used on a PC (under WinNT 4.0, for example), and if so,,) > :would all three buttons be functional?  > N >   AlphaStation systems use the three-button variant of the defacto/standard  >   PC rodent. > K >   I know of no support for IntelliMouse in the X Windows mouse interface.I > L >   Having seen a Microsoft IntelliMouse cause a standard PC to display someN >   seriously bizzare behaviour (no matter which drivers were used), I'd tend L >   to be somewhat hesitatant to use the widget on an OpenVMS system.  (But O >   as it likely won't damage the hardware involved, you can certainly try it.)t > L >   As for what Windows NT might do with a three-button rodent, I'd suggest N >   asking that Windows NT questions in a Windows NT newsgroup.  (I do happen N >   to have a three button mouse connected to a local NT4 box, but the middle N >   button is currently only a "spare" -- its use likely involves finding someM >   sort of custom NT mouse driver for it, a discussion which is WOT (Way Off- >   Topic) for comp.os.vms.) > Z Dunno about MS meeses, but I've always replaced the standard 2 button offering which seemsZ to come with every PeeCee with a Logitech 3 button job. Something in Windows Control PanelX lets you assign double click (and other options) to the middle button. Simply a must forS the main Windows applications - naturally I speak of the REAL Windows applications,iZ FreeCell and Solitaire. Just as naturally, the NT versions of those important applicationsL recognise the double click, as did the W95 ones, brain dead W98 ones do not.    
 Paul Sture Switzerland    ------------------------------  + Date: Fri, 28 Jul 2000 09:38:45 -0400 (EDT)a3 From: lane@DUPHY4.Physics.Drexel.Edu (Charles Lane)o Subject: new release of CRINOIDi8 Message-ID: <000728093832.137@DUPHY4.Physics.Drexel.Edu>  I This is for VMS people who want to do Perl CGI's via the OSU HTTP server.t  F CRINOID is open-source FREE software for running Perl CGIs and serving! them through the OSU HTTP server.o  =     o   CRINOID can provide multiple persistant processes for0A         handling CGI requests for minimum latency.  When combinedyD         with the OSU HTTP's server capability to distribute requestsB         over multiple DECNET nodes, this provides an extreme level3         of responsiveness for VMS CGI applications.   E     o   Perl scripts persistant, cached, and reusable in the multipleeA         server processes for fast response to requests.  Perl andoB         selected Perl modules preloaded to minimize script startup
         time.y  =     o   CGI code for each user handled by separate user-ownediF         processes, so that even unpriviledged users can have CGIs with5         no impact on other user's or system security.7  ;     o   No "255 byte" limitations in HTTP variables; no DCLp1         intermediary between server and Perl CGI.=  G     o   Graphics output directly from Perl, without writing to tempfileS         and copying to server.  E     o   "Standard" Perl CGI code used, although the usual VMS != Unixm         caveats apply.  3     o   Extensive logging and configuration options   C The most recent release is version 0.5-1.  There's been a number of=< bugs fixed and features added over the months since the last
 announcement.4       Some highlights:       o   multihoming supporteH     o   Script timeout control, directory searching and filetype options(     o   Improved configuration procedureB     o   configurable additions to @INC, import/export script %ENV,9     o   Perl scripts with __DATA__ sections now supported'E     o   better error trapping, messages (including user-configurable)h,         and trapping of "exit()" in scripts.E     o   improved script caching and reduced script rerunning overhead   = Source code distribution, documentation, etc. can be found at. http://www.crinoid.com --;  Drexel University       \V                    --Chuck Lanel; ======]---------->--------*------------<-------[=========== .      (215) 895-1545     _/ \  Particle PhysicsM FAX: (215) 895-5934     /\ /~~~~~~~~~~~        lane@duphy4.physics.drexel.eduf   ------------------------------  # Date: Fri, 28 Jul 2000 08:12:55 GMTs/ From: rsmith@nospam.arel.com.au (Russell Smith)t' Subject: OpenVMS on an Alphaserver 2100 1 Message-ID: <bgbg5.1$uf.343@nsw.nnrp.telstra.net>   I I have recently aquired an Alphaserver 2100 4/275.  I am trying, without fB success, to install OpenVMS on it.  I came to me as an NT machine.  ( The model number/part number is 460AP-A9  C When booting from the CD the boot bugchecks with and error loading b PROCESS_MANAGEMENT.EXE ???  M I cannot find anything on the net/wizard that says this machine can only run J< NT, any help with this machine would be greatly appreciated.  
 Russell Smith. Arel Asia Pacificy   ------------------------------  % Date: Fri, 28 Jul 2000 10:37:50 +02000= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>j+ Subject: Re: OpenVMS on an Alphaserver 2100S) Message-ID: <3981465D.64C60FDE@gtech.com>x   Russell Smith wrote:J > I have recently aquired an Alphaserver 2100 4/275.  I am trying, withoutD > success, to install OpenVMS on it.  I came to me as an NT machine. > * > The model number/part number is 460AP-A9 > D > When booting from the CD the boot bugchecks with and error loading > PROCESS_MANAGEMENT.EXE ??? > N > I cannot find anything on the net/wizard that says this machine can only run> > NT, any help with this machine would be greatly appreciated.   An AS 2100 4/275 can run VMS.c   But:   - what firmware version ?    - what VMS version ?   Arne   ------------------------------  % Date: Fri, 28 Jul 2000 11:11:07 +0200k6 From: "Dijk, Jeroen van" <Jeroen.vandijk@getronics.nl>+ Subject: RE: OpenVMS on an Alphaserver 2100nM Message-ID: <2795B75EF003D311801A00A0C906B51187E3F1@cucexec.gbc.getronics.nl>t  I The min VMS version is 6.2, but it also depend on the devices for which =u# a higher VMS version may be needed.gG I don't know about a min for the firmware version, maybe for NT there =r is a min firmware version :)       -----Original Message-----3 From: Arne Vajh=F8j [mailto:arne.vajhoej@gtech.com]w  Sent: vrijdag 28 juli 2000 10:38 To: Info-VAX@MVB.SAIC.COMc+ Subject: Re: OpenVMS on an Alphaserver 2100I     Russell Smith wrote:D > I have recently aquired an Alphaserver 2100 4/275.  I am trying, = withoutrD > success, to install OpenVMS on it.  I came to me as an NT machine. >=20* > The model number/part number is 460AP-A9 >=20D > When booting from the CD the boot bugchecks with and error loading > PROCESS_MANAGEMENT.EXE ??? >=20G > I cannot find anything on the net/wizard that says this machine can =. only run> > NT, any help with this machine would be greatly appreciated.   An AS 2100 4/275 can run VMS.-   But:   - what firmware version ?a   - what VMS version ?   Arne   ------------------------------  % Date: Fri, 28 Jul 2000 05:28:02 -0400s. From: "Vajhoej, Arne" <Arne.Vajhoej@GTECH.COM>+ Subject: RE: OpenVMS on an Alphaserver 2100t> Message-ID: <CD706D1032A7D311BCF60008C75D64FD68AF3F@BELMAIL02>  @ All VMS Alpha versions has a minimum required firmware version !   Arne   -----Original Message-----; From: Dijk, Jeroen van [mailto:Jeroen.vandijk@getronics.nl]  Sent: 28. juli 2000 11:11h* To: 'Arne Vajh=F8j'; Info-VAX@MVB.SAIC.COM+ Subject: RE: OpenVMS on an Alphaserver 2100o    I The min VMS version is 6.2, but it also depend on the devices for which =g am! higher VMS version may be needed.kG I don't know about a min for the firmware version, maybe for NT there =i is a min firmware version :)n       -----Original Message-----3 From: Arne Vajh=F8j [mailto:arne.vajhoej@gtech.com]   Sent: vrijdag 28 juli 2000 10:38 To: Info-VAX@MVB.SAIC.COMe+ Subject: Re: OpenVMS on an Alphaserver 2100e     Russell Smith wrote:D > I have recently aquired an Alphaserver 2100 4/275.  I am trying, = without D > success, to install OpenVMS on it.  I came to me as an NT machine. >=20* > The model number/part number is 460AP-A9 >=20D > When booting from the CD the boot bugchecks with and error loading > PROCESS_MANAGEMENT.EXE ??? >=20G > I cannot find anything on the net/wizard that says this machine can =o only runr> > NT, any help with this machine would be greatly appreciated.   An AS 2100 4/275 can run VMS.o   But:   - what firmware version ?a   - what VMS version ?   Arne   ------------------------------  # Date: Fri, 28 Jul 2000 10:43:36 GMTa/ From: rsmith@nospam.arel.com.au (Russell Smith)t+ Subject: Re: OpenVMS on an Alphaserver 2100m8 Message-ID: <stdg5.20361$c5.56328@newsfeeds.bigpond.com>  H In article <3981465D.64C60FDE@gtech.com>, arne.vajhoej@gtech.com says... >u >Russell Smith wrote:sK >> I have recently aquired an Alphaserver 2100 4/275.  I am trying, withoutlE >> success, to install OpenVMS on it.  I came to me as an NT machine.2 >> -+ >> The model number/part number is 460AP-A9  >> 2E >> When booting from the CD the boot bugchecks with and error loading  >> PROCESS_MANAGEMENT.EXE ???  >> eO >> I cannot find anything on the net/wizard that says this machine can only run9? >> NT, any help with this machine would be greatly appreciated.c >a >An AS 2100 4/275 can run VMS. >  >But:0 >  - what firmware version ? >  - what VMS version ?4 >s  K I have upgraded to the latest firmware provided with the 7.2 kit, and I am  J trying to load 7.2.  I have tried a few different versions from about 6.2  forward with no joy.  O I don't have the exact details of the firmware, I am replying from home at the   moment.   P We have 2 other 2100's in our organisation running VMS, I am not sure though if N they are the same model exactly (ie to part number level, I am waiting from a & reply internally from another site :-(   Thanks for the prompt response.-   Russell.   ------------------------------  % Date: Fri, 28 Jul 2000 15:30:41 +0200d> From: "Jean-Franois Marchal" <jean-francois.marchal@x9000.fr>+ Subject: Re: OpenVMS on an Alphaserver 2100n3 Message-ID: <8ls1qc$1erp$1@s2.feed.news.oleane.net>-  : I think you SHOULD use the ECU DISKETTE for VMS (and unix)5 to fix your eisa bus configuration, then try with VMS    Jean-Franois Marchals X9000 - LYON (FR)i  < "Russell Smith" <rsmith@nospam.arel.com.au> wrote in message+ news:bgbg5.1$uf.343@nsw.nnrp.telstra.net...nJ > I have recently aquired an Alphaserver 2100 4/275.  I am trying, withoutD > success, to install OpenVMS on it.  I came to me as an NT machine. >p* > The model number/part number is 460AP-A9 >sD > When booting from the CD the boot bugchecks with and error loading > PROCESS_MANAGEMENT.EXE ??? >eJ > I cannot find anything on the net/wizard that says this machine can only runs> > NT, any help with this machine would be greatly appreciated. >h > Russell Smith  > Arel Asia Pacific. >n >a   ------------------------------  % Date: Fri, 28 Jul 2000 09:50:23 -0400a% From: "Islandco" <sales@islandco.com>m+ Subject: Re: OpenVMS on an Alphaserver 2100c- Message-ID: <so339q4o0b32@corp.supernews.com>a  0 You might want to check the scsi disk controller  @ I'll bet that if it was running NT, it has an adaptec based card   It don't work with vms  G You will need a Qlogic or NCR/Symbios card that is supported by the SRM?   Davidj  : Russell Smith <rsmith@nospam.arel.com.au> wrote in message+ news:bgbg5.1$uf.343@nsw.nnrp.telstra.net...eJ > I have recently aquired an Alphaserver 2100 4/275.  I am trying, withoutD > success, to install OpenVMS on it.  I came to me as an NT machine. > * > The model number/part number is 460AP-A9 > D > When booting from the CD the boot bugchecks with and error loading > PROCESS_MANAGEMENT.EXE ??? > J > I cannot find anything on the net/wizard that says this machine can only run > > NT, any help with this machine would be greatly appreciated. >o > Russell Smithl > Arel Asia Pacifica >f >    ------------------------------    Date: 28 Jul 2000 11:57:49 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> 2 Subject: Re: Patwhorks Performance and TCP NODELAYH Message-ID: <y4aef2tw4i.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  ' Alan Greig <agreig@my-deja.com> writes:   B > There have been several threads here regarding the abysmal write> > performance of Pathworks Advanced Server on a high speed LAN9 > and I found this to be the case on installing recently.h2 > After an emailed suggestion by Kerry Main to try  > TCPIP>SET PROTOCOL TCP/NODELAYC > the peformance improved *tenfold* and actually beat the NT servere > I was comparing it to.  G If Pathworks doesn't set all its sockets to NODELAY, I have to questiontI the networking competence in general and the TCP competence in particularn of the team producing it.m  G Background: Remember, TCP is basically a protocol supplying a stream of1L bytes. It thus has no concept of a record or buffer or message. If you want M to send a line od text with 80 bytes, say, you could perform one write() withcM all 80 bytes, or you could perform 80 write()s of one byte each, it would all0N be the same (functionally) to TCP. For performance reasons, however, the stackI will usually try to supply as many bytes in one go to the lower layers as J possible (within the limits set by the physical layer(s)) as it can. Thus,I by default TCP will wait a certain time (usually 0.2 seconds) for furthercI bytes to be supplied by the user before actually sending the packet. ThistK makes sense for one mainstream application: telnet. All others usually have L at least a buffer the size the physical layer can take at hand, and can onlyN see a deterioration of performance if NODELAY isn't set. (I don't know who theL idiot was who made DELAY the default. I do know that setting NODELAY was theI first thing I told people to implement, sight unseen, when I heard of the - throughput of our robot-control application.)o  I NODELAYACK is probably a different beast. Windowing and ack-piggybacking eI usually allow delaying of ACKs considerably without impacting performance-J (slightly improving it, actually, because less bandwidth is wasted for theM ACKs). I would have thought that there is no fixed delay, but one that adaptsl/ to the latency and bandwidth of the connection.-  K IIRC, TCP was also designed for slow, long-latency networks with high error9N rates. That leads to a set of tradeoffs that are "wrong" for fast, low-latencyN networks with low error rates (e.g., LANs), such has having slow start. That'sL what you get if you apply a system designed for at most 256 nodes in 30-yearJ old technology and a cooperative research environment to today's internet.   	Jan   ------------------------------  % Date: Fri, 28 Jul 2000 08:24:35 -0700e? From: Mike Price <mike.priceNOmiSPAM@littlewoods.co.uk.invalid>a2 Subject: Re: Patwhorks Performance and TCP NODELAY9 Message-ID: <164b3d0c.a8389851@usw-ex0106-045.remarq.com>=  7 I am a bit confused - I am running UCX 4.2 and the onlye> parameter like NODELAY I can find is the NODELAY_ACK parameter8 but the last post seemed to say there is another NODELAY4 parameter - is it different in UCX 5.0 up or what???@ sorry if I am being stupid but all I have handy until I get back, to the office is HELP and a very old manual!   Mike    ; -----------------------------------------------------------p  7 Got questions?  Get answers over the phone at Keen.com.- Up to 100 minutes free!a http://www.keen.com0   ------------------------------    Date: 28 Jul 2000 17:56:17 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>t2 Subject: Re: Patwhorks Performance and TCP NODELAYH Message-ID: <y47la68d0e.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  A Mike Price <mike.priceNOmiSPAM@littlewoods.co.uk.invalid> writes:   9 > I am a bit confused - I am running UCX 4.2 and the only6@ > parameter like NODELAY I can find is the NODELAY_ACK parameter: > but the last post seemed to say there is another NODELAY6 > parameter - is it different in UCX 5.0 up or what???  M It's quite possible for an implementation to have no globally (i.e., applyingtI to all connections) NODELAY parameter. Individual sochets should have thebM ability to be set to NODELAY. The original poster was using the global methodbK - probably in TCPIP 5.0 - to hack around the apparent failure to do that inw
 Pathworks.   	Jan   ------------------------------  # Date: Fri, 28 Jul 2000 15:06:35 GMTr/ From: "John Nixon" <jorlnixon@worldnet.att.net>s Subject: performance ratingsH Message-ID: <%jhg5.19838$o71.1272546@bgtnsc04-news.ops.worldnet.att.net>  J I have grown to  like the tps ratings that have been published for various
 VAX and AlphalF systems, primarily because when I used them to predict our application performancerF when moving between machines, or when ugrading cpus I would get rather predictable J results.  Once I accepted that VUPS was history, and after the tps ratings were applied0 to most systems available, tps really helped me.  G Are those ratings still being published for new systems?  Where? I alsow liked the way theyJ were published, meaning that they differentiated between different numbers of cpus 	 in a box.o  G I am now trying to compare performance ratings between ES40s, GS80s andy GS140s. L The documents I see do not give tps ratings (except for the older GS140, and they do - not list ratings for various numbers of cpus.b  F Should I assume that the rating are for  fully configured boxes or for minimally configured boxes.PL ie. if a system can take up to eight cpus,  then are the various ratings for
 eight cpusD (or 16 or 4 or whatever) or do the ratings assume a single processorH (eg,  ES40 6/500, 1 to 4 processors,  Specint95 = 27.3)  Is that for one processor or 4 processors.l  = Is there one that is somewhat similar to the old tps ratings.l  H For the ratings that make a "$" comparison (like $/tpmC),  what cost are they using? / TCE,  TCO?  what configuration is that base on?-   ------------------------------  % Date: Fri, 28 Jul 2000 11:24:34 -0400Y+ From: "Main, Kerry" <Kerry.Main@compaq.com>:  Subject: RE: performance ratingsJ Message-ID: <910612C07BCAD1119AF40000F86AF0D8052845CC@kaoexc4.kao.dec.com>   John,o  K <<< I am now trying to compare performance ratings between ES40s, GS80s ande
 GS140s.>>>  rH I suspect one of the following links might be what you are looking for -  A <http://www.compaq.com/alphaserver/performance/perf_by_perf.html>r0 <http://www.compaq.com/alphaserver/a-chart.html>   Regards,  
 Kerry Main Senior Consultant,
 Compaq Canadas Professional Servicesl Voice : 613-592-4660 FAX   : 819-772-7036 Email : kerry.main@compaq.como       -----Original Message-----4 From: John Nixon [mailto:jorlnixon@worldnet.att.net]$ Sent: Friday, July 28, 2000 11:07 AM To: Info-VAX@Mvb.Saic.Coms Subject: performance ratings    J I have grown to  like the tps ratings that have been published for various
 VAX and Alpha F systems, primarily because when I used them to predict our application performancesF when moving between machines, or when ugrading cpus I would get rather predictable J results.  Once I accepted that VUPS was history, and after the tps ratings were applied0 to most systems available, tps really helped me.  G Are those ratings still being published for new systems?  Where? I also. liked the way theyJ were published, meaning that they differentiated between different numbers of cpus 	 in a box.v  G I am now trying to compare performance ratings between ES40s, GS80s and0 GS140s.eL The documents I see do not give tps ratings (except for the older GS140, and they don- not list ratings for various numbers of cpus.   F Should I assume that the rating are for  fully configured boxes or for minimally configured boxes. L ie. if a system can take up to eight cpus,  then are the various ratings for
 eight cpusD (or 16 or 4 or whatever) or do the ratings assume a single processorH (eg,  ES40 6/500, 1 to 4 processors,  Specint95 = 27.3)  Is that for one processor or 4 processors.(  = Is there one that is somewhat similar to the old tps ratings.t  H For the ratings that make a "$" comparison (like $/tpmC),  what cost are they using?t/ TCE,  TCO?  what configuration is that base on?    ------------------------------   Date: 28 Jul 2000 15:25:19 GMT2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman)  Subject: Re: performance ratings6 Message-ID: <8ls8kv$smg$1@mailint03.im.hou.compaq.com>  z In article <%jhg5.19838$o71.1272546@bgtnsc04-news.ops.worldnet.att.net>, "John Nixon" <jorlnixon@worldnet.att.net> writes: ..G :Are those [performance] ratings still being published for new systems?1 ..  G   From the OpenVMS FAQ, one URL with performance comparisons of various/   VAX and Alpha systems is:   @     http://www.digital.com/alphaserver/performance/perf_tps.html  G   Note: having just noticed it, I will remove the (stale) reference to:b  J     http://www.compaq.com/products/workstations/digital/retired/index.html  N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------  # Date: Fri, 28 Jul 2000 12:43:05 GMT:5 From: "Larry D Bohan, Jr" <LBohan@dbc.spam_less..com>eF Subject: Re: Sending a HTML content e-mail from VMS 5.5-2 using MX 4.22 Message-ID: <zXqBOcMNXRSectczHmE3hX917pfC@4ax.com>  7 On Thu, 27 Jul 2000 19:57:38 -0700, "Barry Treahy, Jr."/ <Barry@mmaz.com> wrote:t  H >I'm attempting to e-mail to people an HTML based message which contains2 >among other things URL's to our internal servers. >bF >The general environment is VMS 5.5-2 with MX 4.1.  The receipents areG >using a GUI based reader, Netscape, and at this time, the messages area3 >appearing in complete text, the HTML tags and all.s >cE >Any recommendations?  I've seen in the FAQ using Pine for MIME based = >attachments.  Please don't tell me this is my only option...r > 	 >Regards,- >- >Barry  $ if using perl is an option, perhaps ( look at Mime::Lite, and the CGI modules.     use MIME::Lite;n  7 MIME::Lite->send('smtp', "mail.acme.com", Timeout=>60);D   $file = shift @ARGV; $recipients = shift @ARGV;> $subject = shift @ARGV;                                        $who_from = shift @ARGV;   $fname = $file;v- $fname =~ s/^.*\]//;     # strip VMS dir specn= $fname =~ s/^.*\://;     # strip dir spec for rooted logicalsw  ) $message = "Summary, attached: $fname\n";-  D if (&mail_file( $file, $recipients, $subject, $message, $who_from )) {j+     print "Mailed $file to $recipients.\n";u }c else {e1     print "Error mailing $file to $recipients\n";t     $error_count++;  }i   exit;p  P #-------------------------------------------------------------------------------= # This routine actually mails the file as an attachment using  # base64 encoding.  
 sub mail_filei {f= my ($path, $recipients, $subject, $message, $who_from ) = @_;_   # Create a new message object.  2 my @recipient_list = split( /,\s*/, $recipients );   my $msg = new MIME::Lite From    =>$who_from, To      =>\@recipient_list,a Subject =>$subject,2 Type    =>'TEXT',  Encoding=>'quoted-printable',  Data    =>$message;i   my  $filename = $path;4     $filename =~ s/^.*\]//;     # strip VMS dir specD     $filename =~ s/^.*\://;     # strip dir spec for rooted logicals       if ($path)     {r% #       Attach a file to the message.F  : #       attach $msg Type     =>'application/octet-stream', #       Path     =>$path,v #       Filename =>$filename,a #       Encoding =>'base64',# #       Disposition =>'attachment';n  "         attach $msg Type =>'TEXT',         Path     =>$path,A         Filename =>$filename,e&         Encoding =>'quoted-printable',#         Disposition =>'attachment';e     }h     return $msg->send; }    ------------------------------  % Date: Fri, 28 Jul 2000 06:59:39 +0200r, From: Alan Wildfire <alan.wildfire@sture.ch>1 Subject: Re: Shades of "Nothing sucks like a VAX"-+ Message-ID: <VA.0000008e.19202202@sture.ch>e  > In article <FyDoA6.7Gs@world.std.com>, Terry C. Shannon wrote:B > Newsgroups: comp.os.vms,comp.org.decus,comp.sys.dec,vmsnet.alpha2 > From: "Terry C. Shannon" <shannon@world.std.com>3 > Subject: Re: Shades of "Nothing sucks like a VAX"n% > Date: Thu, 27 Jul 2000 22:31:22 GMTn > " > I guess what can blow, can suck.  F Nothing sucks as hard as Win98 (but you knew that already, didn't you) > = >> After all. it's trademarked by Wildfire Communications....a > $ S'pose I need to change my name too.  9 At least I'm lucky enough that I was not born a McDonald.s ___i
 Alan Wildfirea   ------------------------------  # Date: Fri, 28 Jul 2000 11:08:49 GMTw% From: Alan Greig <agreig@my-deja.com> 1 Subject: Re: Shades of "Nothing sucks like a VAX"r) Message-ID: <8lrpju$8hp$1@nnrp1.deja.com>n  1 In article <39805ABE.7D6DC10E@trailing-edge.com>,V.   Tim Shoppa <shoppa@trailing-edge.com> wrote:? > Seen while driving down the NJ Turnpike last week, on a giant0 > electronic sign: >j > "    THIS IS A TEST OF   " > "      THE VMS SYSTEM    " >eE > Now I'm willing to bet that here "VMS" actually stands for "VehiclecF > Message Signage" or something similar, but if someone wants to prove > me wrong I'm listening :-)  A I have litle faith in the software or people implementing some ofwC these systems here in the UK. A traffic management system which hadnG been in 'learn' mode for some time was switched live on a local holiday = in Dundee. Presumably they picked a local holiday to minimizec@ disruption in case of problems. Unfortunately the data collectedE was so far out of synch for a public holiday that a bug was triggeredtE inverting the traffic light control. Every single light on main roads ( turned red as you approached it. Sigh...  B Local motorway signs here tend to display default messages such asD "Frustration causes accidents", "Watch your speed", "Wear Sunscreen"? and "Floss" (ok, I made these last two up). More often than notu? when they do display traffic information it's for the direction E opposite to the one you're travelling. Signs on approach to the ForthfF Road Bridge (Edinburgh) warning of queuing traffic usually warn of non: existing queues and stay off when there really is a queue.  @ How long before new signs appear saying "Reading pointless signs causes accidents" I don't know.D   --
 Alan Greig    & Sent via Deja.com http://www.deja.com/ Before you buy.S   ------------------------------  % Date: Thu, 27 Jul 2000 16:44:04 -0400 * From: David A Froble <davef@tsoft-inc.com>( Subject: Re: Standalone backups no more?- Message-ID: <39809F14.AF229F30@tsoft-inc.com>4   Joel Millett wrote:. >  > Hi --r > F > I am putting a new, larger hard drive into an Alpha 266 workstation.H > My plan was to create a standalone (bootable) backup tape, boot to it,I > then create an image backup of the old drive onto the tape.  Then I wasuG > going to put the new drive in, boot to my bootable tape, and copy then5 > image down onto the new drive.  Makes sense, right?n > H > Anyway, I was unable to create a standalone backup tape using my trusyD > @sys$update:stabackit command.  VMS told me that "STABACKIT" is no# > longer supported in this version.u > D > Any ideas on how I can get around this?  Is there an easier way to > accomplish my task?e >  > Thanks for any help! > Joel Millett  K Well, first, STABACKIT is VAX only.  For the Alpha, the VMS distribution CDtK provides a decent system to boot to right on the CD.  This is the preferredvO method of running stand-alone backup on the Alpha, when a normal system disk is N not available, ie; as in when restoring your system from tape after losing the system disk.  J The latest VAX CD media also has some of this ability, but nowhere near as versatile as the Alpha version.   L But, what I would do, assuming that your current system disk is working justP fine, is copy straight from that to the new drive.  For example, say your systemP disk is DKA0: and you set the address of the new drive to DKA100:.  You then canK do an image backup, using /IGNORE=INTERLOCK, to the new disk.  Note that nocL other activity should be occuring while this is done, and the networking andP queues could also be shut down.  Then shutdown the system, remove the old drive,I re-address the new drive as DKA0, and boot the system.  Should work fine.j  P However, if you do not already have a second drive, why not just install the new' drive in addition to the current drive.l   Dave   -- g4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Fri, 28 Jul 2000 11:28:35 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>r( Subject: Re: Standalone backups no more?) Message-ID: <39815243.2F301D5F@gtech.com>,   Joel Millett wrote:hF > I am putting a new, larger hard drive into an Alpha 266 workstation.H > My plan was to create a standalone (bootable) backup tape, boot to it,I > then create an image backup of the old drive onto the tape.  Then I wasrG > going to put the new drive in, boot to my bootable tape, and copy thea5 > image down onto the new drive.  Makes sense, right?r   Try:   - boot from VMS CD   - backup disk to tapel   - switch disks   - boot from VMS CD   - backup tape to disks   Arne   ------------------------------  % Date: Fri, 28 Jul 2000 06:59:26 +0200e  From: Paul Sture <paul@sture.ch>* Subject: Re: Stirrings in the undergrowth?+ Message-ID: <VA.0000008b.191fef1c@sture.ch>n  d In article <rdeininger-2707001131460001@user-2ivec5u.dialup.mindspring.com>, Robert Deininger wrote:4 > From: rdeininger@mindspring.com (Robert Deininger) > Newsgroups: comp.os.vmsl, > Subject: Re: Stirrings in the undergrowth?' > Date: Thu, 27 Jul 2000 11:31:46 -0400  > W > In article <00256929.004E1CC1.00@quegw01.btyp>, Steve.Spires@yellowpages.co.uk wrote:  > A > > Contact:   Tel: 3063  -  VSSG, 1st Floor, Bridge Street Plazah > > T > > I wonder the same thing - I have been unable to find  a link to the competition.2 > > All I have found is the winners and the rules. > > E > > Still, should I expect anything more? Perhaps the competition is;C > > 5 > > If you find the competition page, you win a DS10!r > O > I wonder if this is the same contest that ran a few months ago.  The web pageuK > was broken then, and it seemed impossible to actually enter.  This soundsnl > old.  The board of directors hasn't given them permission to delete the web page yet, so it's still there. > J > The webmasters are off being trained in the use of garish colors and newJ > de-humidifier plug-ins.  They'll only be happy when every page contains D > enough non-html that no platform on earth can display it properly. > J Hah! Until yesterday I always thought that the (hawk, spit) M$ website was- far superior to anything Digipaq could offer.   @ Then I tried finding the Service packs for NT, German version...  R Suffice it to say, it is absolutely no wonder that Linux is so popular in Germany.  J > (Yes, I'm being sarcastic.  But it's not nice to lure folks with phantom" > DS10s, so SOMEBODY deserves it.) >e   ___ 
 Paul Sture Switzerlandc   ------------------------------  # Date: Fri, 28 Jul 2000 10:28:00 GMTr% From: Alan Greig <agreig@my-deja.com>  Subject: Re: Such Crap) Message-ID: <8lrn7c$708$1@nnrp1.deja.com>a  & In article <FyDoA4.7Ft@world.std.com>,3   "Terry C. Shannon" <shannon@world.std.com> wrote:- >-E > "Jan Vorbrueggen" <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>  wrote in > >rF > > I do hope they forced CPQ to sign a source escrow agreement before buying.- >-G > Oh, don't worry. They get Michael Capellas' first born son (if he hasu one)G > as escrow. Also 12 of the buildings on the Houston campus. And a buy-s back ofn > the Maynard Mill.e >gE > Now what the hell else do you in your infinite wisdom believe a newi OpenVMS<D > buyer ought to get? I've met with enough of them lately to realize that are" > satisfied with their investment.  B As someone who was very worried about the future of OpenVMS aroundB a year ago (when Compaq dropped Alpha/NT) I can confirm what Terry@ says: Compaq have made very strong statements of ongoing support to worried customers.   C I also have strong reason to believe that I was right to be worriedeC at one point but Compaq looked at the facts, *rejected* phasing outYC VMS and acted accordingly. They've still got some way to go but theIE end of Alpha and/or VMS is no longer on any agenda for the forseeable  future in my opinion.- --
 Alan Greig    & Sent via Deja.com http://www.deja.com/ Before you buy.<   ------------------------------    Date: 28 Jul 2000 16:13:48 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>g Subject: Re: Such CrapH Message-ID: <y4em4ejqar.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  2 "Terry C. Shannon" <shannon@world.std.com> writes:  K > > > [...] ISE (International Securities Exchange) is a NEW Customer [...]eN > > I do hope they forced CPQ to sign a source escrow agreement before buying.L > Oh, don't worry. They get Michael Capellas' first born son (if he has one)N > as escrow. Also 12 of the buildings on the Houston campus. And a buy-back of > the Maynard Mill.   N That would be quite useless if Compaq goes out of business, or decides to drop VMS at some future time.  M > Now what the hell else do you in your infinite wisdom believe a new OpenVMSo > buyer ought to get?v  F Exactly what I said: a source code escrow agreement, paid for by theirK own money, as a form of insurance. I _am_ presuming you know what the word lK "escrow" means. Given your wording above about MC, I am note sure, however.    	Jan   ------------------------------  % Date: Thu, 27 Jul 2000 16:56:12 -0400r* From: David A Froble <davef@tsoft-inc.com> Subject: Re: Updating RMS Fileso- Message-ID: <3980A1EC.AB0DBA06@tsoft-inc.com>o   "Web2.lyon" wrote: >  > Hi,  > G > Does anybody know how to update an RMS indexed file without using thea/ > RTL routines (SYS$UPDATE) under a C program ?w9 > Can I get direct access to data without RMS 'control' ?i >  > Thanks  L Sure, use QIOs.  But you won't be happy, and other users of the file will be reaching for a rope. :-)  P I'm assuming you are talking about the RMS routines.  If instead you are talkingM about the C RTL routines, then yes, you can directly call the RMS routines too# open, read, and write to RMS files.E   Dave   -- e4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Fri, 28 Jul 2000 06:59:39 +0200   From: Paul Sture <paul@sture.ch> Subject: Re: Updating RMS Fileso+ Message-ID: <VA.0000008f.192024d2@sture.ch>n  ; In article <39800005.44636C2E@wanadoo.fr>, Web2.lyon wrote:2* > From: "Web2.lyon" <Web2.lyon@wanadoo.fr> > Newsgroups: comp.os.vmsn > Subject: Updating RMS Filesh' > Date: Thu, 27 Jul 2000 11:25:26 +0200r >  > Hi,A > G > Does anybody know how to update an RMS indexed file without using thei/ > RTL routines (SYS$UPDATE) under a C program ?s9 > Can I get direct access to data without RMS 'control' ?e > E I must ask, WHY do you want to do this? Please state the problem you eH are seeing, and then we can offer an alternative solution. You _really_ F don't want to start writing large portions of stuff that RMS does for  you. ___ 
 Paul Sture Switzerlandh   ------------------------------   Date: 28 Jul 2000 07:29:18 GMT* From: helbig@astro.rug.nl (Phillip Helbig)) Subject: URGENT: VMS keyboard on SUN etc!e. Message-ID: <8lrcoe$aki$1@info.service.rug.nl>  C I am faced with the prospect of having to work somewhere without a nH proper (LK401 etc) keyboard.  I will log in remotely to my VMS machine, I but this is a pain without a proper keyboard.  I don't know what sort of  C support I can expect from the locals there, probably not much.  My oE original plans were to bring along a minimal VMS set up (VAXstation, 0I graphics monitor, keyboard) and essentially use it as an X-terminal, but r& for various reasons this won't happen.  E Assume they just have SUN workstations.  What can I do?  Can I plug a B VT320 into the serial port and then telnet to my machine, with theB keyboard working appropriately?  What sort of physical plug is the serial port on a SUN?  0  B Is there any way I could plug a proper keyboard into the SUN, i.e.H replace the SUN keyboard with mine, without using a terminal through theH serial port?  If I know it will work, I could get a proper keyboard withH a different sort of plug (i.e. not the small phone plug thing), i.e. for a VT510, if my LK401 keyboard.  I PLEASE, someone come out with a VMS laptop (perhaps based on the DS10 L) iF with the possibility of attaching an LK401 to it!!!!!!!!!!!!!!!!!!!!!!  F I'd appreciate replies by email as well since I'll be travelling soon E and just checking my email sporadically and the news group even less. H (I use NEWSRDR with a lot of defined keys on the keypad (making it even G more like VMS MAIL) and have of course inconvenience if I don't have a - proper keyboard.)e   ------------------------------  % Date: Fri, 28 Jul 2000 14:51:39 +0200-& From: Bob Marcan <bob.marcan@aster.si>- Subject: Re: URGENT: VMS keyboard on SUN etc!r( Message-ID: <398181DB.A5C2A6EC@aster.si>  / I'm using LK46W-A2 on Linux, Tru64, VMS and NT.n Maybe will work on SUN too.M   -Bob     Phillip Helbig wrote:n > D > I am faced with the prospect of having to work somewhere without aI > proper (LK401 etc) keyboard.  I will log in remotely to my VMS machine,oJ > but this is a pain without a proper keyboard.  I don't know what sort ofD > support I can expect from the locals there, probably not much.  MyF > original plans were to bring along a minimal VMS set up (VAXstation,J > graphics monitor, keyboard) and essentially use it as an X-terminal, but( > for various reasons this won't happen. > G > Assume they just have SUN workstations.  What can I do?  Can I plug arD > VT320 into the serial port and then telnet to my machine, with theD > keyboard working appropriately?  What sort of physical plug is the > serial port on a SUN?  > D > Is there any way I could plug a proper keyboard into the SUN, i.e.J > replace the SUN keyboard with mine, without using a terminal through theJ > serial port?  If I know it will work, I could get a proper keyboard withJ > a different sort of plug (i.e. not the small phone plug thing), i.e. for  > a VT510, if my LK401 keyboard. > J > PLEASE, someone come out with a VMS laptop (perhaps based on the DS10 L)H > with the possibility of attaching an LK401 to it!!!!!!!!!!!!!!!!!!!!!! > G > I'd appreciate replies by email as well since I'll be travelling soon-G > and just checking my email sporadically and the news group even less.-I > (I use NEWSRDR with a lot of defined keys on the keypad (making it eveneH > more like VMS MAIL) and have of course inconvenience if I don't have a > proper keyboard.)    -- aA -----------------------------------------------------------------c@  Bob Marcan                         email:   bob.marcan@aster.si@  Aster                                tel:    +386 (61) 1894-329@  Nade Ovcakove 1                      fax:    +386 (61) 1894-201@  1000 Ljubljana, Slovenia                    http://www.aster.siA -----------------------------------------------------------------i   ------------------------------  % Date: Fri, 28 Jul 2000 15:00:06 +0200 1 From: "Charles Gilley" <charles.gilley@barco.com>e Subject: Re: Using CMS and NFSI Message-ID: <E472188161D5D311B08700105AF4CAAA01A41515@kuumex03.barco.com>p  ( I don't know if the developer, gee whiz," I wish I could remember his name -3 he deserves the recognition, ever made it to Alpha.n- About the time when Alpha was becoming widelyo0 available, he was in the process of migrating to the PC.-  + Tony Wright <tpw@ngat.com> wrote in messageu< news:04E02AB064E0D211B24E0008C79F6A0E939E40@mail.ngat.com... >cI > Speaking of Doomsday 2000, does anyone know where I can get hold of the0 > VAX/ALPHA versions?eF > I seem to remeber having the same conversation a few years ago to no > avail  >p >. > > -----Original Message-----4 > > From: gilley@nospam.bravewc.com (Charles Gilley)& > > [mailto:gilley@nospam.bravewc.com]+ > > Posted At: Tuesday, July 25, 2000 02:14  > > Posted To: comp.os.vms* > > Conversation: Looking for an old game., > > Subject: Re: Q: Looking for an old game. > >d > >o@ > > Speaking of old games, does anyone know where one might find > > the PC version of D > > Doomsday 2000?  This has to be the classic VAX game of all time. > >c >a   ------------------------------  % Date: Thu, 27 Jul 2000 17:01:43 -04002* From: David A Froble <davef@tsoft-inc.com>C Subject: Re: VMS process manipulate data on SQLServer/NT box : How?X- Message-ID: <3980A337.EA546D2C@tsoft-inc.com>    Stephane Paquin wrote: > : >     Hi! I am looking for an answer to a simple question. > L >     We are currently using an Oracle/VMS VAX to keep our process data. ButF > the old 3100 VAX is nearing the end of its life and I am looking for7 > alternatives. One of them is a NT box with SQLServer.a > H >     Now, my VMS processes on other nodes manipulate data in the OracleJ > server through SQL*net network connection very easily. If  I install theM > NT/SQLServer box, how does a VMS process on another node can manipulate thef@ > data on the server ? What are the tools available to do this ? >  >     Thank you in advance.o >  >     Stphane Paquinl >     Ispat Sidbec  P Why don't you just get a DS10 system?  One answer might be that the Oracle folksN would be looking on over-charging you for this capability.  While the MicroVAXO 3100 is yesterday's low end workgroup class system, and the DS10 is today's lowaN end workgroup class system, some software vendors think that they are entitledG to unlimited access to you bank account when you are just updating (notW upgrading) to new hardware.o   Dave   --  4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Fri, 28 Jul 2000 06:59:23 +0200.  From: Paul Sture <paul@sture.ch>E Subject: Re: WAY OFF TOPIC: help me with the seed of satan pseudo-O/S + Message-ID: <VA.0000008a.191fe4cc@sture.ch>w  N In article <009EDB8A.F45B1B6B@SendSpamHere.ORG>, VAXman- Brian Schenkenberger  wrote: [snip]  G > >Of course, *your* mileage may vary... My remarks about XLNT, which InN > >*personally* find to be a useful and helpful product, were in the spirit ofL > >offering an alternative, like the previous posting re: Acceler8's DCLliteO > >product, to the original problem posting in this thread.  Since I last triedpI > >to contribute constructively to this conversation, I've downloaded thee? > >DCLlite package and given it a spin... it's pretty neat too.b > H > The quality package comes from Accelr8.  The lite version is a freebe,H > parred-down version of their full blown $99.00 package.  Bet you got aG > bigger nose bleed from that other "thing" you're touting.  Next time,oH > learn a little bit about what you're endorsing before you stand up and! > let yourself be counted for it.  > P Right. Having seen your objections about XLNT, I wandered off to their website, R and found that they want 250 (275?) USD for it. More like 500 USD if you take the N package which includes some sort of queue manager (sorry, didn't waste enough 0 time there to do more than get upset with them).  N I first came across them 3.5 years ago, when I was first dealing with NT. The Q idea appealed, but I couldn't find their web address - until 2 weeks after their tR "special offer" of $250 expired. They wanted $395 or similar simply because I had : not ordered sooner (but I couldn't even find them sooner!)  O Ever since then I have regarded them with disdain. AND KERRY MAIN, PLEASE STOP   PROMOTING THEM.d  L After all, there used to be a freeware DCL emulator for DOS available, that H someone had knocked up over a weekend when trapped in some borihg hotel,  ; Compared with $99 from Accelr8, XLNT's price is outrageous.d >  >e ___ 
 Paul Sture Switzerlands   ------------------------------  % Date: Fri, 28 Jul 2000 08:54:39 -0700e1 From: David D Miller <ddmiller@west.raytheon.com>n Subject: WTB: RSX BookF Message-ID: <OF84D5F7AE.FB96DD68-ON0725692A.005702A6@rsc.raytheon.com>   Former RSXers.  B If you have the text,  _RSX -- A Guide for Users_  by John Pieper,J gathering dust on your bookshelves, I'd like to make you an offer.  I need it for training.   dave.c Current RSX guru   ------------------------------   End of INFO-VAX 2000.419 ************************