1 INFO-VAX	Fri, 09 Jun 2000	Volume 2000 : Issue 320       Contents:' Re: Actual vs Theoretical Network Speed / Re: C bashing (was Re: VMS File Caching Futures > Re: Canadian Association of Compaq Users Symposium rescheduled> Re: Canadian Association of Compaq Users Symposium rescheduled> Re: Canadian Association of Compaq Users Symposium rescheduled> Re: Canadian Association of Compaq Users Symposium rescheduled> Re: Canadian Association of Compaq Users Symposium rescheduled Re: Changing password in html? DEC C versions Re: DEC C versions, Re: Detecting memory leak in C with debugger, Re: Detecting memory leak in C with debugger FILESERV@WKU: Updated TQE G Re: Large amount of files in one directory causes very slow	performance C Large amount of files in one directory causes very slow performance G Re: Large amount of files in one directory causes very slow performance G Re: Large amount of files in one directory causes very slow performance G Re: Large amount of files in one directory causes very slow performance G Re: Large amount of files in one directory causes very slow performance G Re: Large amount of files in one directory causes very slow performance - Re: LN15AC toner cartridge compatible to ...?  LPD problems Nic-to-Nic failover  Re: Nic-to-Nic failover  Re: Nic-to-Nic failover  Re: Nic-to-Nic failover  Re: OSU Web Server Problems 0 Re: Qt2.1.0 GUI library now available on OpenVMS! Question on using ASTs for output % Re: Question on using ASTs for output % Re: Question on using ASTs for output   F ----------------------------------------------------------------------  # Date: Thu, 08 Jun 2000 21:27:37 GMT 3 From: "Gord Coulman" <nospam_gcoulman@ccinet.ab.ca> 0 Subject: Re: Actual vs Theoretical Network Speed9 Message-ID: <ddU%4.2804$vT6.331059@news1.telusplanet.net>   G Thanks Jan!  WS_PING is a quite a useful tool.  I ran a few tests using K different target machines and different packet sizes.  With the packet size I all the way up (about 16k), I achieved a best average of 38Mbit/sec and a K peak of 64Mbit/sec.  Normal averages with smaller packet sizes were from 15  to 20Mbit/sec.  H I suppose this means my expensive switching hubs and NICs are doing whatC they're supposed to, as long as the packet sizes are large, anyway.   J I experimented a bit with ttcp as well, but couldn't figure out how to getG useful results from it.  The documentation is of the highly abbreviated : variety (about 20 lines explaining command line switches).   Gord.   J Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> wrote inJ message news:y4g0qorw20.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de...K > Have a look at WS_PING by Ipswitch Inc. (www.ipswitch.com), it contains a K > throughput measurement tool. I'm not totally convined of its accuracy, as G > it has measured the asymptotic speed on my 10 Mbit network to be 11.2  MBit/s.  >  > Jan    ------------------------------  $ Date: Thu, 8 Jun 2000 18:01:55 -0400' From: "Bill Todd" <billtodd@foo.mv.com> 8 Subject: Re: C bashing (was Re: VMS File Caching Futures( Message-ID: <8hp4u9$2od$1@pyrite.mv.net>  E Actually, it's more that I prefer being wise to just being a wise-ass J (especially one incapable of - or at least uninterested in - understandingD the English language rather than reading in his own preconceptions).   - bill  5 David A Froble <davef@tsoft-inc.com> wrote in message ' news:393FB648.5D125C7F@tsoft-inc.com...  > Wayne Sewell wrote:  > > @ > > In article <393EBCF8.E96911AE@tsoft-inc.com>, David A Froble <davef@tsoft-inc.com> writes:  > >  > > [stuff deleted]  > >  > > > I > > > My favorite bitch, and some of you will not agree, is doing a test,  ie; ( if a =D > > > b  vs  if a == b ) and in addition to performing the test, one variable is set L > > > equal to the other.  Things such as this are real easy to do, and real easy to L > > > miss when looking for the error.  (Disclaimer, not sure of the syntax, but you L > > > get the idea.)  Not even going to get into null terminated strings and such.  > > >  > > K > > You need to have Todd the Wise and All Knowing bless your code.  He has  never H > > made such a mistake in all of recorded history and his Ring of Power prevents > > him from ever doing so.  > E > I see he's branching out.  Last year he proclaimed himself the sole  adjudicator J > of who and what software could use VMS.  Must be one hell of a ring. :-) >  > Dave >  > --6 > David Froble                       Tel: 724-529-04506 > Dave Froble Enterprises, Inc.      Fax: 724-529-0596@ > DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com8 > T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  " Date: Thu, 8 Jun 2000 22:25:03 GMT9 From: Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) G Subject: Re: Canadian Association of Compaq Users Symposium rescheduled + Message-ID: <x3x5SVFBYdQa@eisner.decus.org>   o In article <393FA301.41007155@amse.mets.nrcan.gc.ca>, "Dr. Alain Legault" <alain@amse.mets.nrcan.gc.ca> writes:   O >> > It is my opinion that the only way in which to rectify the shortcomings of O >> > what has become of DECUS Canada in its incarnanation as CANACU, is for the ; >> > entire board to resign and for an election to be held.  > ] > I fail to see what purpose this would serve.  First because an election is going on  and we  > have a lot_ > of difficulty finding people who are "ready to step up to the plate".  Second because I don't  > think that] > the board "hasn't done anything".  It certainly failed to communicate the progresses but to 
 > assume that . > nothing has been done is at best simplistic.  = The one resounding theme we have heard over and over again is ? that of all the countries in the world it is Canada whose DECUS ? organization is singularly unable to get it together on the VMS  hobbyist license.   ; If you ask me (from Cambridge, Massachusetts) "How is DECUS < Canada doing?" the answer would be "They still haven't fixed> their VMS hobbyist problem."  In fact, if you ask me how DECUS? chapters around the world are doing, the answer would be "DECUS 3 Canada still can't fix their VMS hobbyist problem."   < Designing a conference and then convincing people to want it< is so much less important than delivering the one thing they$ _do_ want that only you can provide.   ------------------------------  $ Date: Thu, 8 Jun 2000 17:40:17 -0400- From: "Peter Weaver" <peter.weaver@stelco.ca> G Subject: Re: Canadian Association of Compaq Users Symposium rescheduled . Message-ID: <sk04m1qh51127@corp.supernews.com>   > ... = > I fail to see what purpose this would serve.  First because : > an election is going on  and we have a lot of difficulty > ...   F How is there an election going on if we do not know about it? Who getsD to vote? The last document I received from DECUS was a letter askingA for a new logo design, since then I have not received any renewal D notices, any invitations to anything, any notices about elections...  @ Do any other DECUS groups in the world accept members from other. countries? Can I join a DECUS group somewhere?   -- Peter Weaver (former DECUS Canada member)   ------------------------------  # Date: Fri, 09 Jun 2000 01:49:42 GMT 6 From: "Dr Alain Legault" <alain.legault4@sympatico.ca>G Subject: Re: Canadian Association of Compaq Users Symposium rescheduled : Message-ID: <W2Y%4.44026$uw6.950817@news20.bellglobal.com>  = > If you ask me (from Cambridge, Massachusetts) "How is DECUS > > Canada doing?" the answer would be "They still haven't fixed@ > their VMS hobbyist problem."  In fact, if you ask me how DECUSA > chapters around the world are doing, the answer would be "DECUS 5 > Canada still can't fix their VMS hobbyist problem."  > > > Designing a conference and then convincing people to want it> > is so much less important than delivering the one thing they& > _do_ want that only you can provide.   All they need doL is contact Cathy Zigomanis at 416-228-3934, or at cathy.zigomanis@compaq.com to get the ball rolling.  I If you're aware of a problem with members of CANACU getting access to the  license, please let me know.  H If it is JF (though he is not a CANACU member) tell him to call me :-)))  
 Alain Legault    ------------------------------   Date: 8 Jun 2000 20:33:03 -0700 * From: dunnett@mala.bc.ca (Malcolm Dunnett)G Subject: Re: Canadian Association of Compaq Users Symposium rescheduled , Message-ID: <mpHa6dMo7apj@malvm2.mala.bc.ca>  / In article <sk04m1qh51127@corp.supernews.com>,  4      "Peter Weaver" <peter.weaver@stelco.ca> writes: > ? > How is there an election going on if we do not know about it?  > Who gets to vote?   <    Members presumably. Who those may be is anybody's guess. : According to the bylaws I'm a "charter member" ( by virtue= of having been a member in good standing of DECUS Canada when < the new group was formed ), but I've never received anything< from this group, and an email I sent them directly asking if I was a member went unanswered.   = > The last document I received from DECUS was a letter asking  > for a new logo design.  9    That's more than I've ever received. To say this group : has "some communication problems" is understatement of the highest order.  B > Do any other DECUS groups in the world accept members from other0 > countries? Can I join a DECUS group somewhere? >   ?    I keep suggesting the US chapter accept members from Canada,  but they don't seem interested.   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: Thu, 08 Jun 2000 23:57:17 -0400 - From: JF Mezei <jfmezei.spamnot@videotron.ca> G Subject: Re: Canadian Association of Compaq Users Symposium rescheduled , Message-ID: <39406AF6.E214555B@videotron.ca>   Malcolm Dunnett wrote:A >    I keep suggesting the US chapter accept members from Canada, ! > but they don't seem interested.   K Which is why official dissolution of the user group that pretends to be the L successor to DECUS Canada would be helpful since it would allow DECUS-USA to@ accept us as members without stepping on another chapter's toes.  M This might foster formation of LUGs in canada and once they are strong again, / then a new canadian chapter can be re launched.   K Also, since none of the few documents or web site from that user group that K pretends to be the successor to DECUS Canada has bilingual content, (Canada K has 2 official languages and large corporations provide bilingual content), K then the dissolution of the canadian entity and joining of the US one would 4 not make a difference in terms of localised content.   ------------------------------  % Date: Thu, 08 Jun 2000 19:54:14 -0400 % From: JM <vmswiz@geonospamcities.com> ' Subject: Re: Changing password in html? O Message-ID: <3643A52E2339A152.A2B7ECECDB378A47.384655A0EDC3147B@lp.airnews.net>   G > David Pattenden wrote in message <39379943.AD531EC2@wilco-int.com>... 9 > >I'm trying to change my vms password using html forms. / > >Can anybody point me in the right direction?   G Check out my network-wide password changer for VMS, Advanced Server and E Pathworks v5. Runs under OSU and forces users to pick a vms generated E password. Then it uses DECNET to change the password on every node in 
 your network.   ( http://www.geocities.com/vmswiz/vms.html   			*JM*    ------------------------------  % Date: Thu, 08 Jun 2000 19:35:44 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>  Subject: DEC C versions ) Message-ID: <393FD970.48DE2DC5@gtech.com>   1 I thougth DEC C went 5.6-5.7-6.0-6.1, but someone  just mentioned a 6.0A !    ????   Arne   ------------------------------   Date: 8 Jun 2000 16:57:37 PDT T From: Fairfield@SLAC.Stanford.EDU (Ken Fairfield; SLAC: 650-926-2924; FAX: 926-3515) Subject: Re: DEC C versions 3 Message-ID: <Y75a2XAkMd0K@mccdev.slac.stanford.edu>   I In article <393FD970.48DE2DC5@gtech.com>, Arne =?iso-8859-1?Q?Vajh=F8j?=  %     	<arne.vajhoej@gtech.com> writes: 3 > I thougth DEC C went 5.6-5.7-6.0-6.1, but someone  > just mentioned a 6.0A !   H         Well, you also left off 6.2...  The 6.0A is probably/possibly anH     ECO against 6.0.   Similarly,  there  is  an  ECO  for  6.2 (which IH     haven't had time to install) which _could_ generate a 6.2A...but I'dH     expect  a 6.2-xxx instead (where my currently installed version says
     6.2-003).            -Ken --  M  Kenneth H. Fairfield            |  Internet: Fairfield@SLC.Slac.Stanford.Edu :  SLAC, 2575 Sand Hill Rd, MS 46  |  Voice:    650-926-2924:  Menlo Park, CA  94025           |  FAX:      650-926-3515N  -----------------------------------------------------------------------------B  These opinions are mine, not SLAC's, Stanford's, nor the DOE's...   ------------------------------  % Date: Thu, 08 Jun 2000 15:41:16 -0400 - From: JF Mezei <jfmezei.spamnot@videotron.ca> 5 Subject: Re: Detecting memory leak in C with debugger + Message-ID: <393FF6D9.B0EE349@videotron.ca>    Adam Dorosz wrote:P > Try using Heap Analizer. It allows to track memory allocation and check memory
 > contens.S > After compiling and linking your program with /debug run $debug/keep , choos "run 9 > image" in File menu and select "Heap analizer" checkbox   H Thanks for the hint. I had not see the heap analizer thing because I hadD invoked debug by running the image. Will definitely play with that !   ------------------------------   Date: 8 Jun 2000 23:03:56 GMT 2 From: mathog@seqaxp.bio.caltech.edu (David Mathog)5 Subject: Re: Detecting memory leak in C with debugger , Message-ID: <8hp8os$9rt@gap.cco.caltech.edu>  [ In article <393FF6D9.B0EE349@videotron.ca>, JF Mezei <jfmezei.spamnot@videotron.ca> writes:  >Adam Dorosz wrote: Q >> Try using Heap Analizer. It allows to track memory allocation and check memory  >> contens. T >> After compiling and linking your program with /debug run $debug/keep , choos "run: >> image" in File menu and select "Heap analizer" checkbox > I >Thanks for the hint. I had not see the heap analizer thing because I had E >invoked debug by running the image. Will definitely play with that !   E The last time I checked there was an interesting bug in that program. I Apparently there's no wait state in the code that checks the state of the K pull down menus.  So if I pull one down in the heap analyzer window and sit K there thinking, while I'm thinking the program goes to 100% of the CPU and  K all the other windows come to a grinding halt.  Ok to use it, but be quick   with those menu choices!   David Mathog mathog@seqaxp.bio.caltech.edu ? Manager, sequence analysis facility, biology division, Caltech     ------------------------------   Date: 8 Jun 00 13:35:26 -0500 ) From: goathunter@wku.edu (Hunter Goatley) " Subject: FILESERV@WKU: Updated TQE' Message-ID: <IltrUR$biHii@axp1.wku.edu>   ; The following package has been updated on FILESERV@WKU.EDU, ! ftp.wku.edu, and ftp.process.com:       o TQE  (Updated)   ?         TQE V2.1 displays all the Timer Queue Entries currently ?         active on a system.  Great for debugging code that uses E         timers.  Written by Lee Gleason and ported to Alpha by Hunter A         Goatley, this version includes a bug fix submitted by Tom          Chamberlain.    O ------------------------------------------------------------------------------- @ You can get it via the World-Wide Web using a browser and any of the following URLs:   / 	http://www2.wku.edu/www/fileserv/fileserv.html ' 	ftp://ftp.wku.edu/vms/fileserv/tqe.zip / 	ftp://ftp.process.com/wku/vms/fileserv/tqe.zip > 	ftp://ftp.vms.stacken.kth.se/mirrors/wku/vms/fileserv/tqe.zip1 	ftp://ftp.ctrl-c.liu.se/wku/vms/fileserv/tqe.zip " 	ftp://ftp.vsm.com.au/kits/tqe.zip  O ------------------------------------------------------------------------------- ? You can get it via anonymous ftp from ftp.wku.edu; you'll need:   *         [.VMS]UNZIP.EXE or UNZIP.ALPHA_EXE         [.VMS.FILESERV]TQE.ZIP  > The file [.VMS]FILESERV.README contains a brief listing of all- the packages available under [.VMS.FILESERV].   F The files are also available from ftp.process.com under [.WKU.VMS] and4 [.WKU.VMS.FILESERV], from ftp.vms.stacken.kth.se andG ftp.ctrl-c.liu.se under [.MIRRORS.WKU.VMS] and [.WKU.VMS.FILESERV], andh" from ftp.vms.com.au under [.KITS].O -------------------------------------------------------------------------------sB To get it via e-mail, send the following commands in the body of a! mail message to FILESERV@WKU.EDU:s  $ SEND TQE		!Comes as 1 180-block file@ SEND FILESERV_TOOLS     !Needed if you don't have MFTU and UNZIP  D Including the command DIR ALL on a separate line will return a brief< listing of all the packages available from FILESERV@WKU.EDU.     Hunter ------8 Hunter Goatley, Process Software, http://www.process.com8 <goathunter@PROCESS.COM>     http://www2.wku.edu/hunter/   ------------------------------  % Date: Thu, 08 Jun 2000 14:05:33 -0700g* From: Tad Winters <Tad_Winters@corvel.com>P Subject: Re: Large amount of files in one directory causes very slow	performance6 Message-ID: <882568F8.0073A2BE.00@mailhost.corvel.com>  A If files are deleted, it sounds like installing DFU and using itscM DIRECTORY/COMPRESS/TRUNCATE command might be a better solution than trying toc move files between directories.        To:   Info-VAX@Mvb.Saic.Como& cc:    (bcc: Tad Winters/Admin/CorVel) Fax to:iE Subject:  Re: Large amount of files in one directory causes very slow        performanceP      I You didn't mention if files are created AND deleted in that directory ...sD if YES, maybe you should recreate a "new" directory and rename filesB from the old one, and then switch the names ... the .DIR file size can be diminished suche a way.  D Maybe using a logical name pointing on several directory files could? help make creating files quicker ... and retrieval not so long.e  / May need to be tested, but I can't do that now.l   Cordialement Jean-Fran?ois Marchal  X9000 - LYON    # <salvi@freeler.nl> wrote in messagen2 news:8holi1$vuu$1@rubidium.news.lion-access.net... > Hi there,z > L > At this time we have al lot of files, about 25,000 all in one directory on aoJ > VMS 6.2-1H3 2node DSSI cluster with Alpha 4100 machines with each 2Gb of > memory and two CPU's.hI > Both Alpha's are connected to two HSZ50 controllers with 64Mb cache ands host > based shadowing. >sD > The average RMS performance is slowing down as the amount of files	 increasesf > in one directory. F > In the coming months we expect that the amount of files will be near 40,000. 0 > I'm sure that the RMS performance drops again.3 > Is there any tuning possible on RMS / INDEXF.SYS.rK > I was told that when the directoryfile exceeds 128 blocks, performance onoJ > RMS operations will drop very rapidly. (our current directoryfilesize is > about 300 blocks). >nI > Maybe needless to say but filehighwatermarking is already switched off,c and,> > the caching size of the HSZ50 controllers is at maximumsize.K > Of course you can say that we have to decrease the amount of files in onee8 > directory, but that is at the moment no option for us. >@G > If you have any idea how i can avoid these performanceproblems pleasee( > respond to this newsgroup or directly. > Thanxx >s >r5 > For more specific information, feel free to ask me.e >c > Salvi Schrijen > Libertel Network > The Netherlands.  > s.schrijen-nospam-@libertel.nl >e >h >r   ------------------------------  $ Date: Thu, 8 Jun 2000 20:44:42 +0200 From: <salvi@freeler.nl>L Subject: Large amount of files in one directory causes very slow performance8 Message-ID: <8holi1$vuu$1@rubidium.news.lion-access.net>  	 Hi there,t  L At this time we have al lot of files, about 25,000 all in one directory on aH VMS 6.2-1H3 2node DSSI cluster with Alpha 4100 machines with each 2Gb of memory and two CPU's.bL Both Alpha's are connected to two HSZ50 controllers with 64Mb cache and host based shadowing.  L The average RMS performance is slowing down as the amount of files increases in one directory.DL In the coming months we expect that the amount of files will be near 40,000.. I'm sure that the RMS performance drops again.1 Is there any tuning possible on RMS / INDEXF.SYS.iI I was told that when the directoryfile exceeds 128 blocks, performance onkH RMS operations will drop very rapidly. (our current directoryfilesize is about 300 blocks).  K Maybe needless to say but filehighwatermarking is already switched off, andi< the caching size of the HSZ50 controllers is at maximumsize.I Of course you can say that we have to decrease the amount of files in one 6 directory, but that is at the moment no option for us.  E If you have any idea how i can avoid these performanceproblems pleasef& respond to this newsgroup or directly. Thanxx    3 For more specific information, feel free to ask me.-   Salvi Schrijen Libertel Network The Netherlands. s.schrijen-nospam-@libertel.nl   ------------------------------  $ Date: Thu, 8 Jun 2000 21:01:00 +0200> From: "Jean-Franois Marchal" <jean-francois.marchal@x9000.fr>P Subject: Re: Large amount of files in one directory causes very slow performance2 Message-ID: <8hoqe2$ssm$1@s2.feed.news.oleane.net>  I You didn't mention if files are created AND deleted in that directory ...eD if YES, maybe you should recreate a "new" directory and rename filesB from the old one, and then switch the names ... the .DIR file size can be diminished suche a way.  D Maybe using a logical name pointing on several directory files could? help make creating files quicker ... and retrieval not so long.l  / May need to be tested, but I can't do that now.    Cordialement Jean-Franois Marchali X9000 - LYON    # <salvi@freeler.nl> wrote in messageu2 news:8holi1$vuu$1@rubidium.news.lion-access.net... > Hi there,n >eL > At this time we have al lot of files, about 25,000 all in one directory on anJ > VMS 6.2-1H3 2node DSSI cluster with Alpha 4100 machines with each 2Gb of > memory and two CPU's. I > Both Alpha's are connected to two HSZ50 controllers with 64Mb cache andi host > based shadowing. >aD > The average RMS performance is slowing down as the amount of files	 increasesi > in one directory.SF > In the coming months we expect that the amount of files will be near 40,000.c0 > I'm sure that the RMS performance drops again.3 > Is there any tuning possible on RMS / INDEXF.SYS. K > I was told that when the directoryfile exceeds 128 blocks, performance onoJ > RMS operations will drop very rapidly. (our current directoryfilesize is > about 300 blocks). > I > Maybe needless to say but filehighwatermarking is already switched off,  andm> > the caching size of the HSZ50 controllers is at maximumsize.K > Of course you can say that we have to decrease the amount of files in onei8 > directory, but that is at the moment no option for us. > G > If you have any idea how i can avoid these performanceproblems please ( > respond to this newsgroup or directly. > Thanxx >e >r5 > For more specific information, feel free to ask me.n >t > Salvi Schrijen > Libertel Network > The Netherlands.  > s.schrijen-nospam-@libertel.nl >l >i >.   ------------------------------  " Date: Thu, 8 Jun 2000 19:10:03 GMT/ From: "Richard L. Dyson" <rick-dyson@uiowa.edu> P Subject: Re: Large amount of files in one directory causes very slow performance( Message-ID: <393FA93A.7E3EA5E@uiowa.edu>   salvi@freeler.nl wrote:i >  > Hi there,i > N > At this time we have al lot of files, about 25,000 all in one directory on aJ > VMS 6.2-1H3 2node DSSI cluster with Alpha 4100 machines with each 2Gb of > memory and two CPU's.sN > Both Alpha's are connected to two HSZ50 controllers with 64Mb cache and host > based shadowing. > N > The average RMS performance is slowing down as the amount of files increases > in one directory.lN > In the coming months we expect that the amount of files will be near 40,000.0 > I'm sure that the RMS performance drops again.3 > Is there any tuning possible on RMS / INDEXF.SYS.hK > I was told that when the directoryfile exceeds 128 blocks, performance on J > RMS operations will drop very rapidly. (our current directoryfilesize is > about 300 blocks).  @ 	I believe I heard at a DECUS meeting that the only solution youG have (unless you make the dir files smaller than 128 blks) is to update F to OpenVMS v7.2.  I saw a presentation where that was mentioned as one of* the benefits of ODS-5 available in v7.2...  E 	If anyone does have a solution to this performance problem that doesCB not involve updating to v7.2 I know I would love to hear about it!   Rick -- 7H Richard L. Dyson                                    rick-dyson@uiowa.eduH  _   _      _____                http://www-pi.physics.uiowa.edu/~dyson/H | | | |    |_   _|   Systems Analyst                     O: 319/335-1879H | | | | of   | |     The University of Iowa            FAX: 319/335-17536 | \_/ |     _| |_    Department of Physics & Astronomy-  \___/     |_____|   Iowa City, IA 52242-1479a   ------------------------------  % Date: Thu, 08 Jun 2000 16:40:02 -0400m- From: JF Mezei <jfmezei.spamnot@videotron.ca>aP Subject: Re: Large amount of files in one directory causes very slow performance, Message-ID: <3940049B.E0200CBD@videotron.ca>    "Jean-Fran=E7ois Marchal" wrote:F > > The average RMS performance is slowing down as the amount of files > increases  > > in one directory.   J Not the case. RMS kicks in once the file is opened... (Record Management = Services...)6 But the file system (ODS-2 or ODS-5) does slow down...  F > In the coming months we expect that the amount of files will be near	 > 40,000.=  J Do what All-in-1 does for its main share message store. Distribute the fi= lesu in various directories.e   eg: =l  > $DEFINE STORAGE "[store.dir1]","[store.dir2]","store.dir3" etc  J Also, if your application cannot randomize of round-robbin the creation o= f newhJ files, you could simply have a logical name that is updated every hour to=  F point to another directory, and your application simply writes to thatJ logical. This would help distribute the files more evenly across the dire= ctories.  G Furthermore, this technique also would allow you to spread the files ons/ multiple drives which may increase performance.t   ------------------------------  # Date: Thu, 08 Jun 2000 22:55:45 GMT / From: "John Nixon" <jorlnixon@worldnet.att.net>oP Subject: Re: Large amount of files in one directory causes very slow performanceE Message-ID: <RvV%4.1292$ov1.77288@bgtnsc04-news.ops.worldnet.att.net>n  ) The best answer is to upgrade to VMS 7.2.   H This improves the situation in two ways.  It does NOT turn off directory caching whenH the directory file reaches 128 blocks, like previous versions of VMS do.
 Also, it doesgK large block IOs when shuffling directories when files are added or deleted.o  E If you cannot upgrade then you MUST re-distribute files, or live withe decreasing performance, L and the larger the directory gets, the faster the performance will tail off. Eventually, you will+ be getting reports that the system is hung.n  I Of course, you can buy faster disks, and controllers, but even that won'tE scale indefinately.t  K If doesn't sound as though you are deleting files, but if you are, then try  deleting them inL reverse alphabetical order.  I have seen several ways of doing this, but the one I do notL like is where you do an individual delete for each file.  The huge number of image J activations counteracts performance gains.  Our best technique was just to
 delete all! Z files, then all Y files, etc...   K Someone else mentioned DFU to defrag the directory.  I am a big fan of DFU, 	 but there-K are situations where DFU can lose data if acting on an active directory, so= be REAL- careful about that.v  # <salvi@freeler.nl> wrote in messagei2 news:8holi1$vuu$1@rubidium.news.lion-access.net... > Hi there,  >eL > At this time we have al lot of files, about 25,000 all in one directory on a-J > VMS 6.2-1H3 2node DSSI cluster with Alpha 4100 machines with each 2Gb of > memory and two CPU's.FI > Both Alpha's are connected to two HSZ50 controllers with 64Mb cache and  host > based shadowing. >dD > The average RMS performance is slowing down as the amount of files	 increases5 > in one directory. F > In the coming months we expect that the amount of files will be near 40,000. 0 > I'm sure that the RMS performance drops again.3 > Is there any tuning possible on RMS / INDEXF.SYS.sK > I was told that when the directoryfile exceeds 128 blocks, performance oniJ > RMS operations will drop very rapidly. (our current directoryfilesize is > about 300 blocks). > I > Maybe needless to say but filehighwatermarking is already switched off,  andn> > the caching size of the HSZ50 controllers is at maximumsize.K > Of course you can say that we have to decrease the amount of files in one-8 > directory, but that is at the moment no option for us. >5G > If you have any idea how i can avoid these performanceproblems please-( > respond to this newsgroup or directly. > Thanxx >d > 5 > For more specific information, feel free to ask me.a >  > Salvi Schrijen > Libertel Network > The Netherlands.  > s.schrijen-nospam-@libertel.nl >D >c >g   ------------------------------  % Date: Fri, 09 Jun 2000 02:45:53 +0200a= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>tP Subject: Re: Large amount of files in one directory causes very slow performance) Message-ID: <39403E40.44E542AD@gtech.com>p   salvi@freeler.nl wrote: N > At this time we have al lot of files, about 25,000 all in one directory on aJ > VMS 6.2-1H3 2node DSSI cluster with Alpha 4100 machines with each 2Gb of > memory and two CPU's.wN > Both Alpha's are connected to two HSZ50 controllers with 64Mb cache and host > based shadowing. > N > The average RMS performance is slowing down as the amount of files increases > in one directory.oN > In the coming months we expect that the amount of files will be near 40,000.0 > I'm sure that the RMS performance drops again.3 > Is there any tuning possible on RMS / INDEXF.SYS. K > I was told that when the directoryfile exceeds 128 blocks, performance ongJ > RMS operations will drop very rapidly. (our current directoryfilesize is > about 300 blocks). > M > Maybe needless to say but filehighwatermarking is already switched off, andi> > the caching size of the HSZ50 controllers is at maximumsize.K > Of course you can say that we have to decrease the amount of files in oner8 > directory, but that is at the moment no option for us. > G > If you have any idea how i can avoid these performanceproblems pleasel( > respond to this newsgroup or directly.  G The standard advice is either to upgrade to a newer VMS version withoutUB this limit on caching or what is probably still the best solution: createC 10-100 subdirectories under the directories and spread the files in. them..   Arne   ------------------------------  % Date: Thu, 08 Jun 2000 23:59:47 -0400-2 From: "Richard B. Gilbert" <DRAGON@compuserve.com>6 Subject: Re: LN15AC toner cartridge compatible to ...?7 Message-ID: <200006090000_MC2-A818-5D5D@compuserve.com>1  % Message text written by Paul AndersoneF >In article <009EB4CE.6D423224.4@sternwarte.uni-erlangen.de>, Rainer =  0 Sterzer <ai02@sternwarte.uni-erlangen.de> wrote:  H > Who can give me infos to the compatibility of the LN15 to change the =  G > toner cartridge with non-original DEC-hardware. Which other printer =t  0 > (e.g. from HP) needs the same toner cartridge?  B The DIGITAL Laser Printer LN15 and LN15+ have a Fujitsu engine.  =  G However, different brands of printers using the same engine require a =   J different key on the toner cartridge.  So, it is not likely that another =  I brand of cartridge will fit into the LN15.  (A different key physically =u  8 prevents the cartridge from being put into the printer.) <s  F Allowing Compaq/Digital/Genicom to sell the "key" for a stiff premium?   ------------------------------   Date: 9 Jun 2000 02:12:43 GMTv# From: system@niuhep.physics.niu.edu- Subject: LPD problems-+ Message-ID: <8hpjqr$ilj$1@husk.cso.niu.edu>   B Having offered advice on tcpip  printing I am now in need of help.  I LPD is serving a que on a networked printer that has its own LPD. ( a tekh 740)H It also serves a que on an LPS printer, basically just feeding sys$print  N The first I have succeded in getting working the LPS I'm having problems with,4 but that isn't the question I want to ask right now.   tcpip services v 5.0A.  G LPD is having problems recovering from errors.  It takes hours to abortfF an entry in a que.  Unless the sol'n is obvious to somebody I will be I calling compaq about it, but in the meantime I was wondering if I should oH set up multiple queues so that when one gets bogged down the others can B take over.  Is this reasonable?  How hard would this be to set up?
 Any thoughts?    Robert   ------------------------------  $ Date: Thu, 8 Jun 2000 11:55:01 -0700+ From: "Joseph Hom" <Joseph_Hom@hotmail.com>m Subject: Nic-to-Nic failover1 Message-ID: <WXR%4.15$ye2.10210@news.pacbell.net>w  J Has anyone been able to do this. Basically and alpha ds20 running open vmsJ with two nic cards. Is there any way to have a nic-to-nic failover in case. the primary fails. Sort of like 3com's driver.   ------------------------------  # Date: Thu, 08 Jun 2000 20:50:57 GMT 3 From: "Gord Coulman" <nospam_gcoulman@ccinet.ab.ca>>  Subject: Re: Nic-to-Nic failover9 Message-ID: <RGT%4.2801$vT6.329608@news1.telusplanet.net>U  4 Joseph Hom <Joseph_Hom@hotmail.com> wrote in message+ news:WXR%4.15$ye2.10210@news.pacbell.net... L > Has anyone been able to do this. Basically and alpha ds20 running open vmsL > with two nic cards. Is there any way to have a nic-to-nic failover in case0 > the primary fails. Sort of like 3com's driver.  F Side note:  I've had lots of fun with 3Com's driver on NT systems.  AnK undocumented bug sets up one heck of a packet storm if you have two servers I with receive load balancing on the same LAN.  Pegged all the CPUs at 100% G with a switched 100Mbit network.  3Com advised to turn off receive loadi; balancing on one of the machines and the problem went away.t   Gord   ------------------------------  $ Date: Thu, 8 Jun 2000 15:52:54 -0700+ From: "Joseph Hom" <Joseph_Hom@hotmail.com>e  Subject: Re: Nic-to-Nic failover1 Message-ID: <OFV%4.57$ye2.24117@news.pacbell.net>i  H yeah i hear that but my q has to do with vms. has anyone ever heard of a? driver of some sort that will allow nic-to-nic failover in vms? > "Gord Coulman" <nospam_gcoulman@ccinet.ab.ca> wrote in message3 news:RGT%4.2801$vT6.329608@news1.telusplanet.net... 6 > Joseph Hom <Joseph_Hom@hotmail.com> wrote in message- > news:WXR%4.15$ye2.10210@news.pacbell.net... J > > Has anyone been able to do this. Basically and alpha ds20 running open vmsiI > > with two nic cards. Is there any way to have a nic-to-nic failover ins case2 > > the primary fails. Sort of like 3com's driver. >dH > Side note:  I've had lots of fun with 3Com's driver on NT systems.  AnE > undocumented bug sets up one heck of a packet storm if you have twob serverslK > with receive load balancing on the same LAN.  Pegged all the CPUs at 100%hI > with a switched 100Mbit network.  3Com advised to turn off receive loadz= > balancing on one of the machines and the problem went away.  >f > Gord >e >  >e   ------------------------------  % Date: Thu, 08 Jun 2000 20:23:00 -0500d* From: Keith Brown <kbrown780@usfamily.net>  Subject: Re: Nic-to-Nic failover, Message-ID: <394046F4.8C54C771@usfamily.net>   Joseph Hom wrote:, > L > Has anyone been able to do this. Basically and alpha ds20 running open vmsL > with two nic cards. Is there any way to have a nic-to-nic failover in case0 > the primary fails. Sort of like 3com's driver.  ? I have several small Alpha machines running VMS which are usingr= 2 NIC cards but each NIC is connected to a different etherneti= segment.  My understanding is that this is the only supporteda> configuration, which of course is not what you are asking for.   -- d Keith Brownr kbrown780@usfamily.net   ------------------------------  % Date: Tue, 06 Jun 2000 02:26:38 -0700w% From: "Clark L. Zahn" <clzahn@aa.net> $ Subject: Re: OSU Web Server Problems& Message-ID: <393CC3CE.4F54585F@aa.net>  K I don't know enough about VMS systems programming to give more information,aO other than that the webserver process will keep running but stop communicating.p  J If there is someone who would be interested in a short term consulting jobI in the Seattle area send me an email and I'll put you in contact with thewL hiring folks.  They're looking for someone who is an expert with VMS systems? programming and hopefully also is familiar with the OSU server.h       JF Mezei wrote:n >  > "Clark L. Zahn" wrote:; > > We are running into problems with the webserver hangingeJ > > frequently and having to be restarted.  We are writing to a pipe named > > NET_LINK from Fortran. > O > What version of OSU do you have running ? What version of VMS ? You'd need touP > provide a lot more information for anyone to know what is wrong. And you mightB > wish to consider subscribing to the OSU web server mailing list. > M > Have you visited: http://www.er6.eng.ohio-state.edu/WWW/doc/serverinfo.html0   --   *****************************j Clark L. Zahn   clzahn@aa.net    ------------------------------  # Date: Thu, 08 Jun 2000 22:38:32 GMTo2 From: "Zane H. Healy" <healyzh@shell1.aracnet.com>9 Subject: Re: Qt2.1.0 GUI library now available on OpenVMSe6 Message-ID: <IfV%4.355$ru5.106936@typhoon.aracnet.com>  ) Rainer Lehrig <lehrig@t-online.de> wrote:t; > Now that I have ported Qt to OpenVMS it also runs on VMS. > > You can get information about Qt on http://www.trolltech.comD > Qt is the base for the KDesktopEnvironment (KDE) known from Linux.C > I'am in contact with trolltech in order to support Qt for OpenVMSnE > officially. For this we need people willing to test Qt for OpenVMS. F > Qt for OpenVMS is available under http://www.lehrig.de (click button > service)..H > It is a zip file containing a readme.txt and a backup saveset (33 MB).  L Hey, this sounds better than getting Gnome ported!  Personally as a whole, IJ prefer KDE to Gnome.  Although both have some good apps that would be nice to see on OpenVMS.  J I'm FTPing it as I type.  I'll have to see about checking this out tonite.   			Zaner   ------------------------------  % Date: Thu, 08 Jun 2000 11:52:46 -0700y@ From: "Russell E. Owen" <owen@astroNOJNK.washington.edu.invalid>* Subject: Question on using ASTs for output2 Message-ID: <8hoq1u$57mc$1@nntp6.u.washington.edu>  I I have a package that consists of multiple processes and multiple users. 1I Each process can send messages to all users via a message hub. Each user uG has a command interpreter that reads and executes commands and outputs  	 messages.e  ! The present design is as follows:oI - A message hub receives messages in one mailbox and distributes them in  $ to a set of mailboxes, one per user.F - Each command interpreter queues a read of its message mailbox using D SYS$QIO and specifying an AST to run when data is received. The AST B reads writes the message using FORTRAN WRITE to unit 6 and queues F another read. Meanwhile, the interpreter uses a simple loop to read a 0 command (FORTRAN READ on unit 5) and execute it.E - Messaging system errors are reported by writing directly to unit 6 -F (the log file for background processese, the screen for users) rather # than going through the message hub.a  ; Basically this is working, but I'm having several problems:sD - As soon as the user starts typing a command, all output halts. Is I there some easy way to avoid this? (Note: it is only an issue if echo is o- enabled on the terminal, but echo is useful.)n  E - When stress-testing the system, I see occasional errors as follows:G) %FOR-F-RECIO_OPE, recursive I/O operationa   unit 6  file unknown   user PC 00000000D these happen (at least sometimes) in the interpreter, just after an C error has been issued written directly (instead of by the AST); it r& appears to be the AST that is unhappy.  I - (Cosmetic but annoying): when a user types a command it overwrites the fH last output. This is because VMS likes to output stuff as <LF>text<CR>. % Is there some easy way to avoid this?i   So far I have tried:H - Have the interpreter spawn a subprocess to handle output to users. No F improvement, plus it's hard to make sure the subprocess dies when the   user process exits or is killed.H - Use SYS$BRKTHRU to output messages instead of the complex message hub H scheme. (The entire computer is used for this task, so it's really OK). H I used that for years because it was dirt simple, but if a user stopped B reading output then SYS$BRKTHRUW waited to time out, which caused E serious problems for all other users and processes. However, all the .& problems mentioned above were avoided.  H I use high-level reads and writes for user/IO because I can then easily E redirect output. (I like SYS$QIO for many jobs, but it seems complex  F when I/O may be redirected.) I looked at RMS a bit, but it was pretty  intimidating at first glance.   + Any hints or suggests would be appreciated.y  
 -- Russell   ------------------------------  % Date: Thu, 08 Jun 2000 16:49:16 -0400e- From: JF Mezei <jfmezei.spamnot@videotron.ca>z. Subject: Re: Question on using ASTs for output, Message-ID: <394006C4.B4892FE3@videotron.ca>   "Russell E. Owen" wrote:= > Basically this is working, but I'm having several problems:rB > - As soon as the user starts typing a command, all output halts.  J You may need to use QIO for the input/output to terminal. Assuming you are4 using VT terminals, what you could do is as follows:  J Do not use the echo feature for inputs. But capture the input character byK character. If you have a broadcast to place on the screen while the user is== typing a command, you can use VT escape sequences to do this:t& ask VT to save current cursor position1 position cursor where you want message to apprear"
 print messages' ask VT to restore saved cursor positionoK continue with normal input processing (which consists of reading characterss and writing it back.  K The program used by the user can also use the SYS$BRKTHRU system service toiL send a message to itself. That system service has the ability to restore theM text that was entered on the current input line. But it may not be as "clean"a (may cause some scrolling etc).    ------------------------------  $ Date: Thu, 8 Jun 2000 23:43:08 +0200. From: "Jesper Naur" <jesper.naur@post.tele.dk>. Subject: Re: Question on using ASTs for output, Message-ID: <8hp3rn$sd9$1@news.inet.tele.dk>  I Russell E. Owen <owen@astroNOJNK.washington.edu.invalid> wrote in messagen, news:8hoq1u$57mc$1@nntp6.u.washington.edu...G > - When stress-testing the system, I see occasional errors as follows:y+ > %FOR-F-RECIO_OPE, recursive I/O operationh >   unit 6  file unknown >   user PC 00000000E > these happen (at least sometimes) in the interpreter, just after anfD > error has been issued written directly (instead of by the AST); it( > appears to be the AST that is unhappy.  J It is necessary here to realize, that FORTRAN is inherently non-reentrant,F and that goes for FORRTL as well. The problem here is, that an ongoingH FORTRAN write issued by the (non-AST) main program is interrupted by theG AST, which then tries to issue its own FORTRAN write. In certain cases,tL FORRTL detects this and issues the %FOR-F-RECIO_OPE error message. You mightJ want to look into the /REENTRANCY compiler switch, however that appears toJ be supported only after FORTRAN 7.0. If you are on an earlier version, youD must disable AST recognition (using SYS$SETAST) in the main program,J whenever a FORTRAN I/O (write AND read) operation is started and enable itI again afterwards. Take great care to only reenable AST recognition, if itiL was actually enabled before you disabled it (use the SYS$SETAST return value to determine this)!,  L All this indicates, that another real problem is, that your interpreter usesK FORTRAN read as its 'central waiting point'. Since you need AST recognition J at this time, you should avoid FORTRAN read for the user command input andF use SYS$QIO[W] instead. In the first iteration, if you wish to read anH entire user command line using one SYS$QIO[W] operation, this will still; suffer the 'after-first-character' output blocking problem.h  I This is because the VMS terminal driver will block output operations to a G terminal, if there is an outstanding input operation, where one or more G characters have been typed (this is all described in the VMS I/O User'sgG Guide). The solution could then be to ensure, that this situation nevertF occurs: Read one character at a time, but be aware, that this puts theJ burden of line editing and possibly echoing on your program (there you go:+ Implement your own custom terminal driver)!   H You might want to go for a solution, where the code knows whether outputK should be forced using SYS$BRKTHRU (whenever the interpreter is waiting fore. user input) or can be done with FORTRAN write.       Hope to have been of helpn     Jesper Naur    ------------------------------   End of INFO-VAX 2000.320 ************************ this newsgroup or directly. > Thanxx >d > 5 > For more specific information, feel free to ask me.a >  > Salvi Schrijen > Libertel Network > The Netherlands.  > s.schrijen-nospam-@libertel.nl >D >c >g   ------------------------------  % Date: Fri, 09 Jun 2000 02:45:53 +0200a= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>tP Subject:<<< LIST >>> 150 List started. >>> 226 Transfer completed.m" <<< CWD /freewarev40/dsnlink_newA >>> 250 Connected to /disk$misc/decus/freewarev40/dsnlink_new/. <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,89,143 5 >>> 200 Port 89.143 at Host 195.113.19.90 accepted.<
 <<< LIST >>> 150 List started.  >>> 226 Transfer completed.r" <<< CWD /freewarev40/dst-scannerA >>> 250 Connected to /disk$misc/decus/freewarev40/dst-scanner/.t <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,90,146a5 >>> 200 Port 90.146 at Host 195.113.19.90 accepted.w
 <<< LIST >>> 150 List started.e >>> 226 Transfer completed.4 <<< CWD /freewarev40/dtdriver> >>> 250 Connected to /disk$misc/decus/freewarev40/dtdriver/. <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,92,334 >>> 200 Port 92.33 at Host 195.113.19.90 accepted.
 <<< LIST >>> 150 List started.5 >>> 226 Transfer completed./# <<< CWD /freewarev40/dtdriver/srcEB >>> 250 Connected to /disk$misc/decus/freewarev40/dtdriver/src/. <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,93,159 5 >>> 200 Port 93.159 at Host 195.113.19.90 accepted.W
 <<< LIST >>> 150 List started.o >>> 226 Transfer completed.e <<< CWD /freewarev40/dungeon= >>> 250 Connected to /disk$misc/decus/freewarev40/dungeon/.0 <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,95,205i5 >>> 200 Port 95.205 at Host 195.113.19.90 accepted. 
 <<< LIST >>> 150 List started.  >>> 226 Transfer completed.r <<< CWD /freewarev40/dwauth < >>> 250 Connected to /disk$misc/decus/freewarev40/dwauth/. <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,104,55 5 >>> 200 Port 104.55 at Host 195.113.19.90 accepted.e
 <<< LIST >>> 150 List started.> >>> 226 Transfer completed.e! <<< CWD /freewarev40/dwauth/srcY@ >>> 250 Connected to /disk$misc/decus/freewarev40/dwauth/src/. <<< TYPE A >>> 200 Type A ok.  <<< PORT 195,113,19,90,105,1326 >>> 200 Port 105.132 at Host 195.113.19.90 accepted.
 <<< LIST >>> 150 List started.  >>> 226 Transfer completed.r  <<< CWD /freewarev40/dwremterm? >>> 250 Connected to /disk$misc/decus/freewarev40/dwremterm/.  <<< TYPE A >>> 200 Type A ok.  <<< PORT 195,113,19,90,112,1246 >>> 200 Port 112.124 at Host 195.113.19.90 accepted.
 <<< LIST >>> 150 List started.o >>> 226 Transfer completed.e! <<< CWD /freewarev40/dw_newmail @ >>> 250 Connected to /disk$misc/decus/freewarev40/dw_newmail/. <<< TYPE A >>> 200 Type A ok. <<< PORT 195,113,19,90,113,35i5 >>> 200 Port 113.35 at Host 195.113.19.90 accepted. 
 <<< LIST >>> 150 List started.n >>> 226 Transfer completed.a! <<< CWD /freewarev40/emacsv1928@ >>> 250 Connected to /disk$misc/decus/freewarev40/emacsv1928/. <<< TYPE A >>> 200 Type A ok.  <<< PORT 195,113,19,90,122,1176 >>> 200 Port 122.117 at Host 195.113.19.90 accepted.
 <<< LIST >>> 150 List started.n >>> 226 Transfer completed.a% <<< CWD /freewarev40/emacsv1928/cpp D 