1 INFO-VAX	Sun, 24 Apr 2005	Volume 2005 : Issue 228       Contents: Re: Could a PC do this?  Re: Could a PC do this?  RE: Could a PC do this?  Re: Could a PC do this?  Re: Could a PC do this?  Re: Could a PC do this? . Re: Cyrillic fonts with Mozilla and DECwindows Re: DCL File directory browser Re: DCL File directory browser Re: DCL File directory browser Re: Problem with file delete Re: Problem with file delete Re: Problem with file delete Re: Slow Filesystem I/O  RE: Slow Filesystem I/O  Re: Slow Filesystem I/O  Re: Slow Filesystem I/O  Re: Slow Filesystem I/O  Re: Slow Filesystem I/O   Re: VMS to Max OS X file sharing  Re: VMS to Max OS X file sharing  Re: VMS to Max OS X file sharing  F ----------------------------------------------------------------------  % Date: Sun, 24 Apr 2005 07:48:26 +0200 + From: Karsten Nyblad <nospam@nospam.nospam>   Subject: Re: Could a PC do this?= Message-ID: <426b3340$0$67262$157c6196@dreader2.cybercity.dk>    Main, Kerry wrote:G > JF - If you want to know what Intel is really positioning for x86 and B > IA64 (as opposed to making up your own views), check out Intel's' > whitepaper comparing x86 and Itanium:  > J > http://www.intel.com/business/bss/products/server/64-bit_tipping_point.p > df >  > Extract:   > F > "Intel offers two complementary architectural choices that cover theG > full range of 64-bit requirements. One is Intel Itanium architecture, @ > which is designed for the most demanding and business-criticalI > enterprise and technical applications. The other is the family of Intel I > Xeon processor based systems with Intel EM64T. Though not equivalent to H > Itanium architecture in terms of capacity, performance, and RAS, theseF > platforms enable a more gradual migration to 64-bit solutions, sinceJ > they provide native support for existing, legacy 32-bit applications. InI > most enterprise computing environments, both platforms will be needed."   H And what does Intel promise in this text? Nothing, nothing and nothing. I   The Intel site is full of white papers that make it seem like Intel is  E 100% committed to Itanic, but the white papers are worded with great  E care.  Intel never writes that Itanic is an industry standard.  They  I write that it is based on industry standards, and they let partners like  E HP and IDC say that it is.  They even seem pleased with that.  Intel  G does not promise that they will continue to develop Itanic anywhere in  H your quote.  They make it appear like they are backing Itanic 100%, but I when you carefully read their wording, you realize that there is nothing  @ anybody can use against Intel should they decide to drop Itanic.  H Intel's statements are like statements of Bill Clinton:  You had better F be sure to listen carefully, because they are designed to give you an C impression of something, while they are saying something different.    ------------------------------  % Date: Sun, 24 Apr 2005 18:20:09 +0800  From: prep@prep.synonet.com   Subject: Re: Could a PC do this?- Message-ID: <87fyxgqxwm.fsf@prep.synonet.com>   % "Tom Linden" <tom@kednos.com> writes:   C > The fact that there are still a lot of VAXen out there might lead D > you to the conclusion that the Alpha really wasn't that successful  E That is to a large factor due to: drivers that never made the jump to F V6+ versions of VMS, old `obsolete' code that works and a decade laterE people are still trying to get to work on billyshit or unix boxen, SW B vendors who fled DEC, glad hand for the palmer, and left customers	 stranded.      --  < Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda. @                                              West Australia 6076* comp.os.vms,- The Older, Grumpier Slashdot. Raw, Cooked or Well-done, it's all half baked.F EPIC, The Architecture of the future, always has been, always will be.   ------------------------------  % Date: Sun, 24 Apr 2005 09:28:00 -0400 ' From: "Main, Kerry" <kerry.main@hp.com>   Subject: RE: Could a PC do this?R Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB5ECC85@tayexc19.americas.cpqcorp.net>   > -----Original Message-----3 > From: Dave Froble [mailto:davef@tsoft-inc.com]=20  > Sent: April 23, 2005 11:01 PM  > To: Info-VAX@Mvb.Saic.Com " > Subject: Re: Could a PC do this? >=20 > Main, Kerry wrote: >=20@ > > JF - If you want to know what Intel is really positioning=20
 > for x86 and D > > IA64 (as opposed to making up your own views), check out Intel's) > > whitepaper comparing x86 and Itanium:  > >=20 > >=20@ > http://www.intel.com/business/bss/products/server/64-bit_tippi
 ng_point.p > > df > >=20 > > Extract:=20  > >=20H > > "Intel offers two complementary architectural choices that cover the> > > full range of 64-bit requirements. One is Intel Itanium=20 > architecture, B > > which is designed for the most demanding and business-critical> > > enterprise and technical applications. The other is the=20 > family of Intel @ > > Xeon processor based systems with Intel EM64T. Though not=20 > equivalent to B > > Itanium architecture in terms of capacity, performance, and=20 > RAS, theseH > > platforms enable a more gradual migration to 64-bit solutions, since> > > they provide native support for existing, legacy 32-bit=20 > applications. InB > > most enterprise computing environments, both platforms will=20
 > be needed."  >=20F > Sorry Kerry.  What you quote above could be, note, I say 'could',=20H > similar to the EV8 roadmaps and PROMISES and COMMITMENTS concerning=20 > Alpha on June 20, 1999.  >=20A > Actually, Intel is not making commitments and promises, just=20  > 'offers'.=20J > Care to quote for us anything from Intel that PROMISES to keep making=20 > anything?  >=20  
 [Snip ...]  E Dave - Care to quote us anything from any vendor on any platform that 7 "promises" anything will happen in specific timeframes?    That's not reality.=20  D ALL vendors provide directions (roadmaps?) to Customers and AnalystsB based on their best estimates of what they can deliver in specific timeframes.=20  , Now, when certain things happen (engineeringD /legal/marketing/political/stock dives/new mgmt/ new moon/whatever),G these vendors sometimes need to adjust these roadmaps.  That is true of G IBM, Sun, Microsoft, Intel, AMD, Oracle, SAP, etc etc and yes, even HP.   D Here are a few examples from other vendors where "stuff" happens andF things do not always work out the way the vendor announced it to theirG Customers: (and yes, you could argue which vendor was at fault, but the 5 point is, "stuff happened" and the "roadmap" changed)   E "SUN TO DELIVER ENTERPRISE-CLASS SOLARIS FOR INTEL'S MERCED PROCESSOR D Intel to support Sun in moving Solaris to IA-64" - December 16, 1997A http://www.sun.com/smi/Press/sunflash/9712/sunflash.971216.3.html   F "The Road to Cairo" - April 8, 2002 (Cairo supposed to be available in< 1996, then NT5, then Windows 2000 finally delivered in 2000)H http://www.computerworld.com/softwaretopics/os/story/0,10801,69882,00.ht ml   You want "promises"??=20  D Heck, similar to OpenVMS, I would be happy just to see public futureF roadmaps for AIX, Solaris, Linux etc etc. Roadmaps are not meant to beG etched in stone (hence the disclaimers usually found in them), but does H provide some general direction on features and timeframes - based on theG best information that vendor has available at the time the roadmaps are 
 published.   Regards   
 Kerry Main Senior Consultant  HP Services Canada Voice: 613-592-4660  Fax: 613-591-4477  kerryDOTmainAThpDOTcom (remove the DOT's and AT)=20  $ "OpenVMS has always had integrity .. Now, Integrity has OpenVMS .."   ------------------------------  % Date: Sun, 24 Apr 2005 10:03:40 -0400 # From: "John Smith" <a@nonymous.com>   Subject: Re: Could a PC do this?, Message-ID: <7_KdnRXGf_bbOvbfRVn-jQ@igs.net>   Bill Gunshannon wrote:? > In article <4269b90c$0$78285$157c6196@dreader1.cybercity.dk>, / > Karsten Nyblad <nospam@nospam.nospam> writes:  >> Bill Gunshannon wrote: 3 >>> In article <42695693.22166.2FE034B9@localhost>, 8 >>> "Stanley F. Quayle" <squayle@insight.rr.com> writes: >>> 4 >>>> On 22 Apr 2005 at 22:26, Bill Gunshannon wrote: >>>>E >>>>> Nothing has come out yet with enough performance to match, much  >>>>> less emulate the Alpha.  >>>>( >>>> Not yet.  Give it 10 or 15 years... >>>> >>>  >>> G >>> How's that?  5 years for Itanium to die and then 10 more to develop  >>> the new architecture?  >>>  >>> bill >>>  >>G >> 7 years for the Chinese to develop the needed technology and 5 years  >> for them to design the chip.  > A > Since when?  More likely it would be 7 years for someone in the ? > civilized world to develop it and 6 months for the Chinese to  > steal it.     I HP should take the rights they have in Alpha and partner with the Chinese  today.  
 The Benefits: 3 a) VMS would still have a great platform to run on. / b) Linux would have a great platform to run on. L c) Windows would be ported to it in order to serve the large Chinese market.< d) Tru64 customers (those still standing) would have a home.G e) Can you say in your best Carl Sagan voice, "Billions and billions of  Alpha cpu's at low cost"?    ------------------------------  % Date: Sun, 24 Apr 2005 07:48:00 -0700 # From: "Tom Linden" <tom@kednos.com>   Subject: Re: Could a PC do this?( Message-ID: <opspqreay0zgicya@hyrrokkin>  F On Sun, 24 Apr 2005 10:03:40 -0400, John Smith <a@nonymous.com> wrote:  K > HP should take the rights they have in Alpha and partner with the Chinese  > today. > The Benefits: 5 > a) VMS would still have a great platform to run on. 1 > b) Linux would have a great platform to run on. H > c) Windows would be ported to it in order to serve the large Chinese  	 > market. > > d) Tru64 customers (those still standing) would have a home.I > e) Can you say in your best Carl Sagan voice, "Billions and billions of  > Alpha cpu's at low cost"?   I And if they were clear thinking they would also throw in the PC business. E I recently played golf with a former CEO of Intel's biggest rival and G we were talking about HP and IBM and I essentially said the same thing, F to which he responded, what would left?  I said higher margin business, like VMS.  He only looked quizzically at me.   ------------------------------    Date: 24 Apr 2005 09:10:42 -0700* From: "Alan Greig" <greigaln@netscape.net>  Subject: Re: Could a PC do this?C Message-ID: <1114359042.543966.202960@l41g2000cwc.googlegroups.com>    JF Mezei wrote:  > "Main, Kerry" wrote:E > > JF - If you want to know what Intel is really positioning for x86  and D > > IA64 (as opposed to making up your own views), check out Intel's) > > whitepaper comparing x86 and Itanium:  > >  > > H http://www.intel.com/business/bss/products/server/64-bit_tipping_point.p > > df >  > B > Mr Kerry, one needs to distinguish between "official" papers for current E > "politically correct" positions, and all the hints and leaks that a C > company makes over a period to time to indicate future direction.  > E > And when you consider how Intel's position on IA64 has changed over  the A > years, it indicates a very clear trend which you cannot ignore, A > especially since last year, when  Intel accelerated that trend.   + Never mind Intel just look at HP's position 1 http://biz.yahoo.com/bw/050421/206151.html?.v=3D1   C "For example, in a recent four-processor, two-tier SAP=AE Sales and G Distribution (SD) Standard Application Benchmark, results of which were F released today, the dual-core HP ProLiant DL585 supported 1,772 SAP SDF Benchmark users and fully processed 178,000 order line items per hour.D The server supported 755 more users and 76,000 more order line itemsE per hour than the single-core DL585 did in a previous two-tier SAP SD < Standard Application Benchmark, equaling a nearly 75 percent performance improvement.(1)"  E And check the SAP benchmark site for comparisons with Itanic and Xeon G http://www50.sap.com/benchmark/sd2tier.asp which show the new dual core @ Opteron s outperforming Xeons by around 2:1 and roughly matching! Itanium 2 at the same core count.    --=20 
 Alan Greig   ------------------------------  + Date: Sun, 24 Apr 2005 09:30:34 +0000 (UTC) P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply)7 Subject: Re: Cyrillic fonts with Mozilla and DECwindows $ Message-ID: <d4fovq$a79$1@online.de>  5 In article <426AADCA.27872B51@teksavvy.com>, JF Mezei ' <jfmezei.spamnot@teksavvy.com> writes:    1 > Phillip Helbig---remove CLOTHES to reply wrote: E > > is only one size, one shape and one strength.  HTML, in tags like I > > <H1>, <ADDRESS> etc, with "normal" fonts results in a different size, ( > > shape or strength than the default.  > E > You can't really know what font/size H1 will want because it can be  > specified in a style sheet.   E Yes, I realise that it isn't specified specifically in the HTML page  ? itself, but there has to be some algorithm which determines it.    ------------------------------  % Date: Sun, 24 Apr 2005 01:41:32 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> ' Subject: Re: DCL File directory browser , Message-ID: <426B3165.A93E5C92@teksavvy.com>   stinehelferw wrote:  > ? > Has anyone found a way to use DCL to browse a file directory?   D The reverse can be done quite easily for the decw$ fileview. You canF browse directories and then execute dcl on one or more selected files.   ------------------------------  % Date: Sun, 24 Apr 2005 03:24:53 -0400 + From: "stinehelferw" <stinehelferw@cox.net> ' Subject: Re: DCL File directory browser , Message-ID: <dRHae.4858$lz1.4596@lakeread01>   Using a DCL menu. G After selecting function/operation need to find the file to operate on. H Something like the File/Open dialog on a text editor.  Just something to find the file and select it.    C "Peter 'EPLAN' LANGSTOEGER" <peter@langstoeger.at> wrote in message & news:426b5cd2$1@news.langstoeger.at...= > In article <RFEae.4846$lz1.3958@lakeread01>, "stinehelferw"  <stinehelferw@cox.net> writes:@ > >Has anyone found a way to use DCL to browse a file directory? >  > What is browsing for you ? > List files ? > Checking attributes ?  > J > >I'm thinking C is a much better idea, but thought I'd check as I have a menu9 > >of DCL utilities that this would be incorporated into.  > K > If you need file attributes of open files, then you need to code in C (or > > other languages), because F$FILE_ATTRIBUTES can't show them. >  > --   > Peter "EPLAN" LANGSTOEGER ' > Network and OpenVMS system specialist  > E-mail  peter@langstoeger.atH > A-1030 VIENNA  AUSTRIA              I'm not a pessimist, I'm a realist   ------------------------------  % Date: Sun, 24 Apr 2005 05:22:42 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> ' Subject: Re: DCL File directory browser , Message-ID: <426B652D.EEC8095E@teksavvy.com>   stinehelferw wrote:  >  > Using a DCL menu. I > After selecting function/operation need to find the file to operate on. J > Something like the File/Open dialog on a text editor.  Just something to > find the file and select it.  C To find a list of files, you can do a DCL look with F$SEARCH("*.*") ' until F$SEARCH returns an empty string.   E DCL doesn't have arrays per say, but you can still concuct something:    $numfiles=0  $loop1: " $	FILE_'numfiles = F$SEARCH("*.*")/ $	if FILE_'numfiles .eqs. "" then GOTO endloop1  $	temp = file_'numfiles - $	write sys$output "''numfiles'      ''temp'"  $	numfiles = numfiles + 1  $goto loop1  $! $ 
 $endloop1:/ $inquire select "Enter number of file you want"  $the_file = file_'select $edit 'the_file   E What the above does is assign the variable FILE_0 with the first file 1 name, FILE_1 with the second file name and so on.   C If you have decwindows, fileview is really what you want to use. It H gives you file list and you can define multiple actions that can be doneG on each file (right click pops a menu of actions on that file, based on  the file type).    ------------------------------  % Date: Sun, 24 Apr 2005 18:28:03 +0800  From: prep@prep.synonet.com % Subject: Re: Problem with file delete - Message-ID: <87br84qxjg.fsf@prep.synonet.com>   G "@no$pam.comcast.net" <""thomas.simpson1\"@no$pam.comcast.net"> writes:   = > We intermittently get a failure that appears to be due to a E > significant delay after a large file is deleted and the free blocks D > info on the disk is updated.  Of course, by the time I log back inC > (after hours) the free space on the target disk looks normal.  Is  > this a know problem or issue?   6 > The file that is deleted is very large (29M blocks).  $ OK, that is kind of large I guess...  B > The configuration is a 2 node ES40 cluster (using memory channelB > adapters) running VMS 7.3-1.  The disk is on an EMC SAN and is a. > single disk consisting of 4 physical drives.  8 > No there is no other activity on the disk at the time.  E > The procedure is trying to delete a file from a disk and copy a new B > file back to the same disk to replace it.  The copy fails due to
 > disk space.   D > Here is a fragment of the log file that shows the problem.  Due toC > seeing this problem before, I had put in code to display the free A > space on the disk before the delete, a 10 second wait after the D > delete and display the free space again after the delete.  It also > issues a SHOW DEVICE command.   ) > Free disk space before DELETE: 24762064 H > %DELETE-I-FILDEL, DISK$DATA2:[CAP.DATA]HISTORY.DAT;1 deleted (29117872	 > blocks) = > Free disk space after DELETE: 24762064  File Size: 26315496 L > %COPY-E-OPENOUT, error opening DISK$DATA2:[CAP.DATA]HISTORY.DAT; as output= > -RMS-F-FUL, device full (insufficient space for allocation) H > %COPY-W-NOTCOPIED, DISK$CNVWORK:[CONVERTWORK1]HISTORY.DAT;1 not copied  F >   A CRITICAL ERROR HAS OCCURRED WHILE CONVERTING HISTORY...PROCEDURE > ABORTING....) >   ERROR: COPY ERROR  STATUS: %X106710A2   G > Device                  Device           Error    Volume         Free I >   Name                   Status           Count     Label        Blocks H > $1$DGA415:    (OSIJX1)  Mounted              0  DATA2         24762064  H Ah, any chance the file is still open? If so, the delete will not reallyH start untill the file is closed. Till then, it is only marked for delete  and the directory entry removed.  I If that is the case, it should also screw up a COPY/OVERLAY as well. Find ; the process that has the file open and talk to it nicely ;)    --  < Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda. @                                              West Australia 6076* comp.os.vms,- The Older, Grumpier Slashdot. Raw, Cooked or Well-done, it's all half baked.F EPIC, The Architecture of the future, always has been, always will be.   ------------------------------  % Date: Sun, 24 Apr 2005 07:33:14 -0400 + From: Ken Robinson <kenrbnsn1@patmedia.net> % Subject: Re: Problem with file delete A Message-ID: <6.2.3.0.2.20050424073104.02422588@mail.patmedia.net>   J At 09:06 PM 4/23/2005, @no$pam.comcast.net <"@broadbandsupport.net, wrote:= >We intermittently get a failure that appears to be due to a  E >significant delay after a large file is deleted and the free blocks  D >info on the disk is updated.  Of course, by the time I log back in C >(after hours) the free space on the target disk looks normal.  Is   >this a know problem or issue?5 >The file that is deleted is very large (29M blocks).   B Do you have highwater marking enabled on the volume? (show dev/fu C disk) That causes a significant delay. Do you have erase-on-delete  # enabled for the file? (dir/fu file)    Ken    ------------------------------  % Date: Sun, 24 Apr 2005 10:31:52 -0400 E From: "@no$pam.comcast.net" <""thomas.simpson1\"@no$pam.comcast.net"> % Subject: Re: Problem with file delete 0 Message-ID: <OuadnVn3rY9KMPbfRVn-tQ@comcast.com>   prep@prep.synonet.com wrote:  I > "@no$pam.comcast.net" <""thomas.simpson1\"@no$pam.comcast.net"> writes:  >  > = >>We intermittently get a failure that appears to be due to a E >>significant delay after a large file is deleted and the free blocks D >>info on the disk is updated.  Of course, by the time I log back inC >>(after hours) the free space on the target disk looks normal.  Is  >>this a know problem or issue?  >  > 6 >>The file that is deleted is very large (29M blocks). >  > & > OK, that is kind of large I guess... >  > B >>The configuration is a 2 node ES40 cluster (using memory channelB >>adapters) running VMS 7.3-1.  The disk is on an EMC SAN and is a. >>single disk consisting of 4 physical drives. >  > 8 >>No there is no other activity on the disk at the time. >  > E >>The procedure is trying to delete a file from a disk and copy a new B >>file back to the same disk to replace it.  The copy fails due to
 >>disk space.  >  > D >>Here is a fragment of the log file that shows the problem.  Due toC >>seeing this problem before, I had put in code to display the free A >>space on the disk before the delete, a 10 second wait after the D >>delete and display the free space again after the delete.  It also >>issues a SHOW DEVICE command.  >  > ) >>Free disk space before DELETE: 24762064 H >>%DELETE-I-FILDEL, DISK$DATA2:[CAP.DATA]HISTORY.DAT;1 deleted (29117872	 >>blocks) = >>Free disk space after DELETE: 24762064  File Size: 26315496 L >>%COPY-E-OPENOUT, error opening DISK$DATA2:[CAP.DATA]HISTORY.DAT; as output= >>-RMS-F-FUL, device full (insufficient space for allocation) H >>%COPY-W-NOTCOPIED, DISK$CNVWORK:[CONVERTWORK1]HISTORY.DAT;1 not copied >  > F >>  A CRITICAL ERROR HAS OCCURRED WHILE CONVERTING HISTORY...PROCEDURE >>ABORTING....) >>  ERROR: COPY ERROR  STATUS: %X106710A2  >  > G >>Device                  Device           Error    Volume         Free I >>  Name                   Status           Count     Label        Blocks H >>$1$DGA415:    (OSIJX1)  Mounted              0  DATA2         24762064 >  > J > Ah, any chance the file is still open? If so, the delete will not reallyJ > start untill the file is closed. Till then, it is only marked for delete" > and the directory entry removed. > K > If that is the case, it should also screw up a COPY/OVERLAY as well. Find = > the process that has the file open and talk to it nicely ;)  >   F Thanks for the tip.  After re-examining the log file a little closer, F (I'm embarrassed to say) I can see that a rogue process still had the F file open.  I had assumed the CONVERT would fail if this happened (it  usually does).  D To be safe I also re-checked the high-water marking and erase-after   delete settings and they are OK.  . Thanks to all that responded with suggestions.   Regards, Tom    ------------------------------  # Date: Sun, 24 Apr 2005 11:28:39 GMT ! From: Nigel Barker <nigel@hp.com>   Subject: Re: Slow Filesystem I/O8 Message-ID: <ui0n61hvoj3asl9t8qejd7frpahgg43cpk@4ax.com>  M On Sat, 23 Apr 2005 14:09:30 -0400, Bill Todd <billtodd@metrocast.net> wrote:   L >When OpenVMS says it has completed an I/O it really has written it to disk. > C >The same, of course, is true for Windows:  it's only higher-level  J >operations such as at the file level where the differences which I noted   >above start to become relevant.  K I think that you are playing with semantics here. Compare & contrast a copy N operation to a USB memory stick. If I yank it out as soon as I get the commandM prompt back there is a good chance that under Windows my files are corrupt or ( incomplete. This is not true of OpenVMS.   -- Nigel Barker Live from the sunny Cote d'Azur    ------------------------------  % Date: Sun, 24 Apr 2005 09:58:42 -0400 ' From: "Main, Kerry" <kerry.main@hp.com>   Subject: RE: Slow Filesystem I/OR Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB5ECC86@tayexc19.americas.cpqcorp.net>   > -----Original Message-----1 > From: ksidibaba [mailto:ksidibaba@gmail.com]=20  > Sent: April 23, 2005 11:12 AM  > To: Info-VAX@Mvb.Saic.Com  > Subject: Slow Filesystem I/O >=20G > Working for the first time on an openVMS system, I found the file I/O D > performance to be extremly slow compared to a wintel notebook. TheE > openVMS platform I am using is a DS15, running openVMS 7.3, patched F > with the recommended patches from HP in order to run Java 1.4.2. TheH > file I/O operations involved consist in writing a log file from a JavaG > program (i.e. an ODS 5 file). Writing approximately 80MB of data to a H > file takes about 60-80s, on a notebook about 8s (4s on a standard PC).C > The same happens when copying the same file to the same directory F > under a different name using dd from GNV (in order to eliminate JavaA > as the possible origin of the problems). Because I'm still very G > unexperienced with the platform, I suspect some of the quotas, amount F > of buffers, etc, to be not well configured. The system consists of aC > DS15 with 1GB of RAM, 1 GHz Alpha CPU, 2x36 GB 15K 320 SCSI disks D > (which I would normally expect to be pretty quick when it comes to > disk I/O operations). ' > Any hint/help would very appreciated. 	 > Thanks,  >=20 > Karim  >=20   Karim,  $ A few suggestions for consideration:  D - VMS V7.3 is 4 years old right now (SSB April 2001), so it is worthC getting a more up to date version of OpenVMS than this. Perhaps you D actually have a more recent version i.e. a "-x" release of VMS V7.3?C Ideally, you will have VMS V7.3-2 as it had some fairly significant 3 performance enhancements over previous versions.=20   E As an example, a recent post on comp.os.vms. stated their own testing H showing a 2-1 performance increase with OpenVMS V7.3-2 over V7.3-1 using simple copy command.  D - As someone else mentioned, ensure you follow the doc's on quota's, system parameters etc.H http://h18012.www1.hp.com/java/documentation/1.4.2/ovms/docs/user_guide. html#processquotas  = - check out and read the following OpenVMS Java tuning guide: H http://h71000.www7.hp.com/ebusiness/OptimizingSDKGuide/OptimizingSDKGuid e.html  E - as per the doc's, please ensure that you run the following to check  your Java environment:3 SYS$COMMON:[JAVA$14x.COM]JAVA$CHECK_ENVIRONMENT.COM   B - this may or may not be of use in your testing, but turn off fileD highwater marking (security parameter that is on by default) on each disk under test: $ set volume $1$duax /nohigh  H - when testing and comparing write IO's to UNIX/Windows, set OpenVMS for? write-back as per the following (assumes VMS V7.3-1 or higher): ) (reference 5.11, but specifically 5.11.3) < http://h71000.www7.hp.com/doc/731FINAL/6657/6657pro_007.html   Regards   
 Kerry Main Senior Consultant  HP Services Canada Voice: 613-592-4660  Fax: 613-591-4477  kerryDOTmainAThpDOTcom (remove the DOT's and AT)=20  $ "OpenVMS has always had integrity .. Now, Integrity has OpenVMS .."   ------------------------------  % Date: Sun, 24 Apr 2005 10:04:31 -0400 2 From: Bob Willard <BobwBSGS@TrashThis.comcast.net>  Subject: Re: Slow Filesystem I/O0 Message-ID: <_6Sdna3DBYn8OvbfRVn-3Q@comcast.com>   Dave Froble wrote: > ksidibaba wrote: > H >> Working for the first time on an openVMS system, I found the file I/OE >> performance to be extremly slow compared to a wintel notebook. The F >> openVMS platform I am using is a DS15, running openVMS 7.3, patchedG >> with the recommended patches from HP in order to run Java 1.4.2. The I >> file I/O operations involved consist in writing a log file from a Java H >> program (i.e. an ODS 5 file). Writing approximately 80MB of data to aI >> file takes about 60-80s, on a notebook about 8s (4s on a standard PC).  >  > E > Nigel and Bill explained some of the issues.  I'd just ask, do you  D > really believe that any system does a substained write to disk of  > 20MB/sec?  I don't.  >   D I'm afraid your expectations of sequential R/W performance are a bitF behind the times.  The 74GB Raptor on this PC sustains ~70 MB/s on its9 outer bands, and I've measured that under XP with HDtach.  --   Cheers, Bob    ------------------------------    Date: 24 Apr 2005 08:21:08 -0700* From: "Alan Greig" <greigaln@netscape.net>  Subject: Re: Slow Filesystem I/OC Message-ID: <1114356068.260432.193010@o13g2000cwo.googlegroups.com>    Bill Todd wrote:   > G > Any operating system at all serious about its future would have fixed   D > such default behavior a decade or more ago:  looking like an utter slug= > when compared with the default behavior on Windows or Unix, 
 especiallyC > while providing no stronger guarantees that the data has actually  made. > it onto the disk as compensation, is absurd.  ; Recall Bill that VMS did fix this problem a decade ago with C Spiralog/TNFS/Files-64. Sure it had problems but rather than fix or B re-implement the programmers were sacked during downsizing and theD product dropped. With that file system in place a simple write of anC 80MB file could indeed complete almost instantly. I ran a full-feed E news-server (ANU News) with Spiralog field-test for some time and met ? some of the developers. The project was based at DEC Livingston * (Edinburgh) and I still have the t-shirts.  ) >   When OpenVMS says it has completed an ) > > I/O it really has written it to disk.  > C > The same, of course, is true for Windows:  it's only higher-level D > operations such as at the file level where the differences which I noted ! > above start to become relevant.  >  > - bill   ------------------------------  + Date: Sun, 24 Apr 2005 16:59:49 +0000 (UTC) % From: Dan Foster <usenet@evilphb.org>   Subject: Re: Slow Filesystem I/O6 Message-ID: <slrnd6nk9v.78k.usenet@gaia.roc2.gblx.net>  m In article <1114356068.260432.193010@o13g2000cwo.googlegroups.com>, Alan Greig <greigaln@netscape.net> wrote:  > = > Recall Bill that VMS did fix this problem a decade ago with E > Spiralog/TNFS/Files-64. Sure it had problems but rather than fix or D > re-implement the programmers were sacked during downsizing and theF > product dropped. With that file system in place a simple write of anE > 80MB file could indeed complete almost instantly. I ran a full-feed G > news-server (ANU News) with Spiralog field-test for some time and met A > some of the developers. The project was based at DEC Livingston , > (Edinburgh) and I still have the t-shirts.  C That's a shame (Spiralog). Wonder why then-DEC dropped it if it was C pretty good? There's some rumours of boundary conditions presenting F significant engineering work making it unpalatable to develop further?  J Note to self: a quick primer on filesystem writes -- three major policies:  @ 	a) Write-through caching -- app writes to fs; fs doesn't return@ 	   until all data blocks has been successfully written to disk.  B 	   Safest but quite slow for obvious reasons. Default on OpenVMS.  A 	b) Write-back caching -- app writes to fs; fs returns as soon as @ 	   data is in cache, even if cache has not yet flushed to disk.  A 	   Very unsafe -- very good performance but if interrupted (e.g. = 	   power failure) at wrong time, filesystem data corruption.   > 	c) Write-behind caching -- app writes to fs; fs reorders data> 	   as necessary in order to guarantee that data will never be= 	   corrupted even if interrupted mid-write, and also retains 1 	   performance properties of write-back caching.   D 	   Provides performance of write-back with safety of write-through.  E The performance you mentioned sounded like it used write-back or some > variant. So being curious, I checked a paper on Spiralog from:  A http://research.compaq.com/wrl/DECarchives/DTJ/DTJM04/DTJM04P8.PS   7 ...written by some of the key engineers on the project.   F I didn't have a good PostScript viewer so I ran it through ps2pdf on aH non-OpenVMS machine with excellent results and then viewed the resulting	 PDF file.   C Apparently, Spiralog utilized write-behind caching which gained the G benefits of write-back performance and had the safety of write-through.   G I liked the engineering considerations put into it such as knobs to set B policy on a per-file basis as well as at other layers (e.g. RMS on	 OpenVMS).   E Did any remnants of the work make it into any component of subsequent  OpenVMS releases?    -Dan   ------------------------------    Date: 24 Apr 2005 19:33:23 +0100K From: pmoreau@ath.cena.fr (Patrick MOREAU, CENA Athis, Tel: 01.69.57.68.40)   Subject: Re: Slow Filesystem I/O! Message-ID: <00fEa1awx1f$@sinead>   h In article <EemdnerXZcfWTvffRVn-1A@metrocastcablevision.com>, Bill Todd <billtodd@metrocast.net> writes: > Dave Froble wrote: >> ksidibaba wrote:2 >> oI >>> Working for the first time on an openVMS system, I found the file I/O F >>> performance to be extremly slow compared to a wintel notebook. TheG >>> openVMS platform I am using is a DS15, running openVMS 7.3, patchedoH >>> with the recommended patches from HP in order to run Java 1.4.2. TheJ >>> file I/O operations involved consist in writing a log file from a JavaI >>> program (i.e. an ODS 5 file). Writing approximately 80MB of data to aeJ >>> file takes about 60-80s, on a notebook about 8s (4s on a standard PC). >> / >> eF >> Nigel and Bill explained some of the issues.  I'd just ask, do you E >> really believe that any system does a substained write to disk of s >> 20MB/sec? > G > Sure.  It's a log file - strictly sequentially written.  If the file uI > isn't significantly fragmented and there are no  periodic log 'forces'  I > (i.e., if the application can just let the cache to flush out the data dI > on its own) steady-state write bandwidth, even to a single (reasonably aH > contemporary) disk (as long as no one else is competing for it or for I > other system resources) could easily maintain 20 MB/sec (even on VMS - mC > but you'd have to tell RMS to multi-buffer and use write-behind).p >  > - bill  M You can try to include into your login.com file (or the login.com file of thee0 application's account the three following lines:   $ set rms/seq/block=127/buf=8c $ set rms/ind/buf=20 $ set rms/extent=4096n  M I use them as my defaults and write performance is really better (write at at 1 least 10 Mb/sec on Ultra SCSI disks with a DS10).s  G Bill is right, rms default values are stupid. They were OK for VAX 780 m5 (1 mips) with 4 Mb memory and disks at 500 Kb/sec ...    Patrickr --O ===============================================================================sN pmoreau@ath.cena.fr  (CENA/SDER) ______      ___   _          (Patrick MOREAU)4 moreau_p@decus.fr (DECUS)       / /   /     / /|  /|J CENA/SDER/Athis-Mons France    / /___/     / / | / |   __   __   __   __  N BP 205                        / /         / /  |/  |  |  | |__| |__  |__| |  |N 94542 ORLY AEROGARE CEDEX    / /   ::    / /       |  |__| | \  |__  |  | |__|N http://www.ath.cena.fr/~pmoreau/              http://membres.lycos.fr/pmoreau/O ===============================================================================e   ------------------------------  # Date: Sun, 24 Apr 2005 11:17:45 GMTe" From:   VAXman-  @SendSpamHere.ORG) Subject: Re: VMS to Max OS X file sharingg0 Message-ID: <00A42C26.2CEF526D@SendSpamHere.ORG>  i In article <6vednbzcjvQsSPffRVn-vA@comcast.com>, Brad Hamilton <braMdhamAiltPonS@comMcasAt.nPeSt> writes:.! >VAXman- @SendSpamHere.ORG wrote:tl >> In article <PtCdnW0s3M2d2cPfRVn-sw@comcast.com>, Brad Hamilton <brMadAhaPmiSlton@coMmcAasPt.Snet> writes: ><snip>s >>>oL >>>Could someone elaborate on potential file sharing problems between a Mac F >>>running the latest OS X and VMS?  I currently use Samba on my home L >>>network, and my wife is considering purchase of a Mac laptop.  I'd still H >>>like to have her access her file shares on VMS using Samba, but I'll ( >>>consider other options, if I have to. >> g >> aK >> My wife has been using my son's 12" Powerbook to perpare the taxes.  ShehL >> uses a drive I created with Samba to keep the tax files (it's on a shadowK >> set on my VMS box and it gets backed up).  The only issue I seem to haveiK >> is that Samba makes the drives 20MBs and I can't figure out how to allowC >> a bigger drive. >> > > >Any luck fixing that with the "latest-'n-greatest" Samba/VMS?  I I've been busy and haven't had a chance to check it out.  Is there a pre- 2 compiled newer version of Samba ready for install?     --  K VAXman- A Bored Certified VMS Kernel Mode Hacker   VAXman(at)TMESIS(dot)COMe            e5   "Well my son, life is like a beanstalk, isn't it?" u   ------------------------------  % Date: Sun, 24 Apr 2005 08:18:35 -0400 6 From: Brad Hamilton <braMdhamAiltPonS@comMcasAt.nPeSt>) Subject: Re: VMS to Max OS X file sharingd0 Message-ID: <n-ydndpAetUGE_bfRVn-iA@comcast.com>    VAXman- @SendSpamHere.ORG wrote:k > In article <6vednbzcjvQsSPffRVn-vA@comcast.com>, Brad Hamilton <braMdhamAiltPonS@comMcasAt.nPeSt> writes:p > " >>VAXman- @SendSpamHere.ORG wrote: >> <snip>K >>>My wife has been using my son's 12" Powerbook to perpare the taxes.  SheFL >>>uses a drive I created with Samba to keep the tax files (it's on a shadowK >>>set on my VMS box and it gets backed up).  The only issue I seem to havepK >>>is that Samba makes the drives 20MBs and I can't figure out how to allown >>>a bigger drive. >>>4 >>? >>Any luck fixing that with the "latest-'n-greatest" Samba/VMS?  >  > K > I've been busy and haven't had a chance to check it out.  Is there a pre-o4 > compiled newer version of Samba ready for install? >  >e  H Unfortunately, no - it needs compiling.  The compile did go cleanly for * me, so it shouldn't take much babysitting.   ------------------------------  # Date: Sun, 24 Apr 2005 14:33:37 GMT ; From: "Jeffrey H. Coffield" <jeffrey@digitalsynergyinc.com>h) Subject: Re: VMS to Max OS X file sharing 8 Message-ID: <57Oae.45$Gd7.34@newssvr21.news.prodigy.com>   Brad Hamilton wrote:   > J > Wife purchased iBook today - transferring all her "important" files via D > Samba/VMS, without too many problems.  Setting up SMB was not too J > difficult, but when the iBook boots, it seems to take a few minutes for B > the network shares to show up.  I haven't figured out the magic 3 > incantations to make the process speedier.    :-)   G The main problem I've seen is with InDesign on the Mac. It will always  F save a file the first time on a samba share but won't save the file a H second or successive time. Also NetBeans has some problem with the date H on a file so that it keeps saying the file has been modified externally  and should be reloaded.s   ------------------------------   End of INFO-VAX 2005.228 ************************