<<< TEMBEL::OFFICE$DISK:[NOTES$LIBRARY]LANGUAGES_AND_TOOLS.NOTE;1 >>> -< The Languages and Tools SIG >- ================================================================================ Note 1.0 Welcome ... No replies BIGUN::KEOGH "I feel better than James Brown" 5 lines 6-AUG-1991 00:03 -------------------------------------------------------------------------------- Welcome to VAX NOTES This conference is for discussions relating to The Languages and Tools SIG  ================================================================================ Note 2.0 LaTeX DVIALW font problem 3 replies TEMBEL::BEISSELC15 "Craig Beissel Coles Myer Ltd." 23 lines 20-AUG-1991 08:47 -------------------------------------------------------------------------------- We have just purchased a few DEClaser 2250s and have started using the DVIALW driver to produce PostScript files from our LaTeX documents. The DVIALW driver comes from the LaTeX tape of a couple of years ago. I have a set of font files created by METAFONT containing the company logo in a number of variations and sizes. (I'll tell you that story another time). These fonts all print correctly on an LN03 using DVILN3 to produce an .LN3 file. My problem is that »some« sizes of the font do not print correctly when using DVIALW. The problem is probaly not the printer, as an LN03R prints exactly the same. DVIALW seems to be getting confused when it is downloading charaters to the printer. HELP! All suggestions gratefully accepted. ================================================================================ Note 2.1 LaTeX DVIALW font problem 1 of 3 VAXNET::VETTE "Sheep are Natural Blondes" 8 lines 20-AUG-1991 11:08 -< Apple Laser Writer? >- -------------------------------------------------------------------------------- Craig, I presume that DVIALW is DVI to Apple Laser Writer? Is there a more generic DVI to PostScript© available that may give you more reliable output? Lindsay Digital (NZ) ================================================================================ Note 2.2 LaTeX DVIALW font problem 2 of 3 TEMBEL::BEISSELC15 "Craig Beissel Coles Myer Ltd." 7 lines 20-AUG-1991 11:18 -< DVIALW is from the DVI drivers family >- -------------------------------------------------------------------------------- DVIALW is the Apple Laser Write driver. It is from the DVI driver family that includes LN03, LA75 etc. I do not know of another Postscript driver, but I have not had an exhaustive look through the Library LaTeX distribution. If anyone knows of anther driver I would be happy to try it. ================================================================================ Note 2.3 LaTeX DVIALW font problem 3 of 3 VAXNET::FRYA22 "Andrew Fry Praxa Ltd." 8 lines 21-AUG-1991 11:34 -< Use DVI2LN3, DVI2PS or PSDVI >- -------------------------------------------------------------------------------- Two ideas: 1. Use DVI2LN3 and send the LN3 file to the ANSI queue for your Postscript printer. Perhaps not the most efficient, but it works. 2. Use the more robust DVI to PS drivers available on the LaTeX distribution, such as DVI2PS and PSDVI. I stopped using DVIALW some time ago because of other such flaky problems.  ================================================================================ Note 3.0 ORACLE perfromance 6 replies TEMBEL::HARRY27 "The Ghost Who Walks ... " 24 lines 20-AUG-1991 14:13 -------------------------------------------------------------------------------- I manage a VAX/VMS site that runs ORACLE (V6). We recently went through an exercise of buying more memory due to page faults rates getting too high. We run 14 databases, the largest of which is 250MB. The problem we has was estimating accurately how much memory we should buy to really make the system fly (CPU does not seem to be the bottleneck) Memory consuption fell into 2 camps - users working sets and the databases themesleves. To cut out a long story we eventually decided to add 32MB and suck it and see. i.e. increase memory allocation to the users and databases until we get acceptable fault rates and hope we dont run out of memory first. Hardly satisfactory however, has anyone had the same experience ? I am thinking of presenting a paper next year since it seems an area that everyone would love to know more about. What do you think ? ================================================================================ Note 3.1 ORACLE perfromance 1 of 6 TEMBEL::ROBINSONM31 13 lines 20-AUG-1991 15:15 -< Oracle performance? On a VAX? >- -------------------------------------------------------------------------------- I manage a VAX/VMS site running ORACLE V6 (Oracle Financials). We run a 6510 with 64mb of memory for about 30 users. Memory is not a problem, the bottle necks are disk I/O and CPU. Our working sets sizings are large (4mb per user) but we are still experimenting with optimum page faulting rates versus user memory requirements. Indications are about 2mb may be the minimum. The system does not "fly". The databases use up enormous amounts of global pages/sections. We are forced to restrict the environment to as few active databases as possible but spread out separate DB tablespace on as many spindles to reduce I/O load. Oracle's estimate for memory is approx 2mb per user plus an overhead for the database server etc of 4-6mb (depending on the size of the SGA) and variable amount per active database. ================================================================================ Note 3.2 ORACLE perfromance 2 of 6 VAXNET::SUZUKIA25 14 lines 22-AUG-1991 10:51 -< Oracle problems and V6.2 >- -------------------------------------------------------------------------------- I manage a VAX/VMS site running ORACLE V6 and have also recently increased user working set sizes to reduce page faulting. This did have the desired effect in regards to page faulting, but we are now running out of some other resource which we haven't been able to pinpoint as yet. The symptom is that ORACLE will decide to hibernate users for sometimes up to 45mins - 1 hr. This is despite the fact that CPU and memory utilizations appear normal. Has anyone else encountered this problem? Another question - has anyone had any experience with ORACLE V6.2 on a cluster yet? We are currently trying it out on a DSSI cluster with VMS5.4-2 and it runs like a dog. Apparently, ORACLE does not use DLM in the conventional sense (DLM goes via the DSSI bus in VMS 5.4-2) but uses ethernet instead. This doesn't seem to make much sense to me. ================================================================================ Note 3.3 ORACLE perfromance 3 of 6 VAXNET::GARNERR30 "Whale Oil Beef Hooked" 15 lines 22-AUG-1991 12:28 -< Oracle Performance is an oxymoron, isn't it ? >- -------------------------------------------------------------------------------- > The problem we has > was estimating accurately how much memory we should buy to really make > the system fly (CPU does not seem to be the bottleneck) The best answer (without buying the memory) will come from modeling system performance. Either get DEC in to use DECcp, or use VPA's modeling subsystem. The former is the better, but the latter is cheaper. Another possibility is to try graphing working set size for the heaviest page faulting processes against hard fault rate. Project these up, add this to per-user requirements, build a spreadsheet, and you've got an answer. This is not as good as modeling, but its quick and cheap, and better than going in blind. ================================================================================ Note 3.4 ORACLE perfromance 4 of 6 TEMBEL::HARRY27 "The Ghost Who Walks ... " 10 lines 22-AUG-1991 15:25 -< Locking ? >- -------------------------------------------------------------------------------- The problem may be due to a user taking out an exclusive lock on a table that others users need (eg security related) We had a problem with similar symptoms that turned out to be a crook RF31 controller. We are running a DSSI cluster but not sharing databases or exe code and have no prblems at all. ================================================================================ Note 3.5 ORACLE perfromance 5 of 6 VAXNET::ROBINSONM31 12 lines 23-AUG-1991 08:54 -< Oracle Performance - DECcp doesn't help >- -------------------------------------------------------------------------------- Oracle is designed in a way that does not allow DECcp (or VPA etc) to measure true performance (eg response times). Modelling system performance is very difficult to impossible because of the back-end database facilities used. DECcp cannot relate the interactive part to the back-end part and gives totally incorrect picture of system performance. It can give a general feeling about how the system is running but cannot relate who is actually consuming resources. Database commits are the main bottleneck (via I/O to disks) and that load cannot be related back to the individual interactive user. The only way to calculate user response times I have found was to write a nasty little DCL procedure to measure when buffered I/O finishes and then the direct I/O finishes. ================================================================================ Note 3.6 ORACLE perfromance 6 of 6 TEMBEL::HARRY27 "The Ghost Who Walks ... " 5 lines 23-AUG-1991 11:15 -< Its not I/O >- -------------------------------------------------------------------------------- Strangely enough our bottleneck is not I/O - probably due to DSSI and the number of disks we have. The real problem is memory and how to tune the databases.  ================================================================================ Note 4.0 L&T SIG Elections No replies VAXNET::FRYA22 "Andrew Fry Praxa Ltd." 5 lines 22-AUG-1991 17:36 -------------------------------------------------------------------------------- Andrew Fry (FRYA22) has been returned at the SIG Chair and Alistair Grant (GRANTA15) has been returned as the SIG Librarian. At this stage, there are no plans to fill the position of SIG Editor.  ================================================================================ Note 5.0 Bugs in COBOL 4.4 2 replies VAXNET::FRYA22 "Andrew Fry Praxa Ltd." 5 lines 22-AUG-1991 18:01 -------------------------------------------------------------------------------- Has anyone else suffered the severe bugs that we found in this release. Some of the are unresolved. Most of the bugs occur using the /INSTRUCTION_SET=NODECIMAL_STRING qualifer. We have received 2 patches to date. Thery manifest themselves as incorrect values after a COMPUTE statement. ================================================================================ Note 5.1 Bugs in COBOL 4.4 1 of 2 PENCEL::DEMO 1 line 23-AUG-1991 09:00 -< CDD/Administrator demo >- -------------------------------------------------------------------------------- Come and see us at the CASE stand for a look at the forthcoming product... ================================================================================ Note 5.2 Bugs in COBOL 4.4 2 of 2 VAXNET::THORNEW29 11 lines 23-AUG-1991 10:37 -< Cobol V4.4 Pioneer is using. >- -------------------------------------------------------------------------------- We are in the process of converting our large Cobol Library to Cobol V4.4. We have not started debugging yet. Since I am not a Cobol Programmer and indeed have not written a programme for 26 years I do not know whether the bug to which you refer manifests itself at Compile of Debuggibg Stage. I have not heard complaints at Compile stage so may be I have yet to hear of the problems. Thank you for advising the problem. I will check and call you in the next two weeks.  ================================================================================ Note 6.0 VAXSET INQUIRY No replies VAXNET::THORNEW29 24 lines 23-AUG-1991 10:45 -------------------------------------------------------------------------------- We are the owners of the Vaxset product. We are using CMS and MMS in a far better way today than ever before. We are moving into using PCA and SCA. However I would like to use DTM - but even after all these years it is at the back of a very long list of requests. When I say "I" I mean my I would like my support staff to use it to thoroughly test changes to long extant programs. Does anybody have any experience of using DTM ? From attempts in the past, the staff complained about slowness and the lack of diskspace. I have addressed both of these now but have yet to re-implement it. In the meantime I have to retrain the staff as those trained have left. We have a three VAX LAVC, ie two MicroVax II and One Vaxstation. We do have plenty of disk space, I obtained some cheap second hand gear. Any comments appreciated, as I would like my Supervisory staff to learn through DECUS of other people's thoughts. Thank You Bill Thorne.