Info-Vax-REQUEST@KL.SRI.COM.UUCP (05/20/87)
Info-Vax Digest Wednesday, 20 May 1987 Volume 0 : Issue 29 Today's Topics: Variants of VMS V4.5 RE: What is DECUS C and Where do I get it. RE: Is an LAVC worth it (Answer: YES YES YES) Backup to WORM/Laser-disks RE: INDEXF.SYS extensions LAVC Ethernet Performance - II Request for VMS FINGER Program BIG backup tapes Re: TPU virtual memory loss problem (Info-VAX 12-May-1987) ---------------------------------------------------------------------- Date: Wed, 20 May 87 04:40:48 PDT From: nagy%warner.hepnet@lbl.arpa Subject: Variants of VMS V4.5 >Where is the difference between 4.5, 4.5a, 4.5b. V4.5A includes support for Local Area VAXClusters. V4.5B (I've forgotten) V4.5C includes support for the VAXStation-2000 and MicroVAX-2000. >Is it true that 4.5a supersedes any existing MicroVMS distribution? Does not seem so. We use VMS V4.5A on our LAVCs (all MicroVAXes). The variants introduced by V4.5A..C are merged back into a single copy of VMS with V4.6. I believe, at DECUS, it was mentioned that V4.6 would be out in July. =Frank Nagy =Fermilab Research Division EED/Controls =FNAL::NAGY.HEPNET or NAGY@FNAL.Bitnet ------------------------------ Date: Wed, 20 May 87 04:54:14 PDT From: nagy%warner.hepnet@lbl.arpa Subject: RE: What is DECUS C and Where do I get it. >I have heard that a C compiler is available through DECUS. Does it run >under VAX/VMS? If so, will it run on 4.* systems? If so, where can I get >it and how much does it cost? The DECUS C compiler will run under VMS, but in compatibility mode. So you need the Digital VAX-11/RSX layered product to run it. The DECUS C compiler produces PDP-11 code which will run on the VAX in compatibility mode. If you want a native mode C compiler for the VAX, buy VAX C from Digital (its approximately the same cost as VAX FORTRAN). VAX C is a very good product from my experience with it. The next release, V2.3, will include many of the features in the ANSI C standard awaiting acceptance (such as function prototypes, volative and const storage class modifiers, etc.). DECUS C will run under VMS V4.x systems. The C System Language tape is a full 2400' magnetic tape; I believe that the DECUS Library cost for the tape is about $150. Of course, if you have to buy VAX-11/RSX also, your costs will be considerably higher. A version of the C System appeared on a recent (1986?) RSX SIG tape also. The DECUS C System tape includes much more than just the C compiler and its libraries. There are libraries written in C which can be compiled under VAX C and used on the VAX (in fact there is one library explicitly for VAX C; this includes the getredirection() routine to simulate UNIX- like I/O redirection on the command line). There are also many utilities written in C which can be compiled under VAX C and used on the VAX in native mode. In short, the DECUS C System is a bargain for PDP-11 and VAX users. =Frank Nagy =Fermilab Research Division EED/Controls =FNAL::NAGY.HEPNET or NAGY@FNAL.Bitnet ------------------------------ Date: Wed, 20 May 87 05:08:47 PDT From: nagy%warner.hepnet@lbl.arpa Subject: RE: Is an LAVC worth it (Answer: YES YES YES) > How good is LAVC and are there any snags? >DEC have told me the advantages, but are there disadvantages too? DEC have >warned that the Ethernet load would be high, so we'd need a bridge to keep that >off the rest of our net, but what about performance and load on the 'boot >member'? And are there any less obvious problems? We do not want to go >through all the trouble of converting to LAVC then find we have to revert! My group is running 2 LAVCs. The WARNER LAVC is used for development and consists of two MicroVAXes. Each uVAX is equipped with 16 MB of memory. The boot member has a KDA50 with two RA81s (system disk and user disk). The satellite member has an RQDX3 controller and an RD53 currently used only for local, secondary paging and swapping files. The DISNEY LAVC is the control system (in development) operational system and consists of three uVAXes, all of which have 16 MB of memory. The RA60 system disk and RA60 user disk are connected to KDA50s in the boot member and satellite #1 (the boot member mounts the system disk, satellite #1 the user disk which it also uses for local paging/swapping files). Satellite #2 has an RQDX3/RD53 only used for local, secondary paging/swapping. Due to (1) the size of the LAN system here at Fermilab and (2) the expected very high control system traffic between the DISNEY LAVC and the Front End MicroVAXes, both LAVCs sit on their own Ethernet segments with LANBridges connecting them to the main LAN systems (Fermilab current has about 70+/-15 VAXes and MicroVAXes (and a few PDP-11s) connected to a multi-segment Ethernet LAN). All terminal connections to both LAVCs are via DECServer-200s or SET HOST. Future plans call for addition of VAXStations to WARNER for programmer workstations and to DISNEY as consoles for the control system. Performance of the LAVC has been outstanding. The LAVC provides all the features (except performance and redundancy) of the CI-cluster (I used to manage a moderately large CI-cluster). Especially for the boot and other core members (any satellite serving "large" disks to the other members of the LAVC), its probably important to take advantage of the low price of MicroVAX memory and put a full 16 MB on such systems. In general, one can then tune the system to "burn memory for better performance". We use large WSEXTENTs, increase file XQP caches, and increase size of the modified page list (really cuts down on paging I/O). We have been quite pleased with the systems; in fact the WARNER cluster has been up for 30 days as I write this now. One piece of information I remember having quoted to me from the LAVC performance session given at Spring DECUS (similar to session given last Fall) is that a MicroVAX-II boot node saturates at 85 I/Os per second (CPU limited). Thats like beating 2.5 RA81s to death! And you can spread the load by having other systems serve public user and data disks to the LAVC. =Frank Nagy =Fermilab Research Division EED/Controls =FNAL::NAGY.HEPNET or NAGY@FNAL.Bitnet ------------------------------ Date: Wed, 20 May 87 14:43:48 +0200 From: Tore Haraldsen USE/UiO <x_haraldsen%use.uio.uninett@NTA-VAX.ARPA> Subject: Backup to WORM/Laser-disks We are contemplating our backup situation to decide if we need to buy another tape-station, or if we should go for other kinds of backup media. Can somebody expand on laser-disks for backup (what harware is around, performance, appx. cost etc)? Dec or other vendors? Tore Haraldsen University of Oslo Norway ------------------------------ Date: Wed, 20 May 87 08:55 EDT From: "Ronald A. Jarrell" <JARRELLRA%vtcs1.cs.vt.edu@RELAY.CS.NET> Subject: RE: INDEXF.SYS extensions Yes there is a problem with cluster. Yes, DEC knows about it. Yes, I'm not surprised that DEC UK doesn't, because it took the file systems specialist 20 minutes of talking to people to figure out who knew what the bug was. They were so busy finding and fixing it they apparently forgot to document it or some such. YES THERE IS A FIX FOR IT. It is fixed in 4.6. Since it's a little hard to install 4.6 without the release kit, DEC HAS A PATCH. Call them if you have software support and ask for it. Mine is coming in a week or so and when I get it I'll post the ECO/SPR numbers. Patch IS version specific, since it affects F11BXQP. The description as given to me by DEC is - "If one node decides that INDEXF.SYS needs to be extended, it will do that, and everyone will start putting all headers in the extended area. If in the small window of time between it doing that, and it writing it's cache back to disk it goes down, then all headers created during that time are lost, since one of the other nodes will now extend indexf on it's own." Ron Jarrell Va Tech JarrellRA@vtcs1.cs.vt.edu ------------------------------ Date: Wed, 20 May 87 06:46:28 PDT From: nagy%warner.hepnet@lbl.arpa Subject: LAVC Ethernet Performance - II (Left out of my previous LAVC message) On Ethernet loading... Fermilab has a LAN Traffic Monitor from Digital. When we initially installed our 2-node WARNER LAVC, it was connected directly to the main Ethernet segment. This segment includes a number of busy VAXes, including a very large CI-cluster and the WAN routers (Fermilab is a major player on HEPnet). The LTM showed that this segment was only about 3-5% utilitized normally with peaks in the 11-15% region. The WARNER LAVC was usually in the top 10 Ethernet users but often not the topmost user. Its biggest effect on the Ethernet was when booting when, for a short time, the nodes were sending lots of messages (Boot Me! Boot Me!). As we more than triple the number of nodes in WARNER with workstations this might change, but WARNER is already on its own segment so the effect should be relatively minor. On the other hand (to set the scale), we have succeeded in scaring the network managers here. A test of our distributed control system code (uses DECnet) with 4 MicroVAXes and 16 processes (4 data producers and 12-16 consumers) with a 20 Hz data repetition rate (every 50 msec) and long messages (about 300 items resulting in messages of about 3 Kbytes) succeeded in raising Ethernet utilization to about 40%. On the other hand this is, somewhat, an unrealistically large load for the planned normal operating state of our control system. The point I'm making here is that the MicroVAXes and Ethernet were able to quite easily absorb this load (the LAVC traffic was, incomparison, insignificant). =Frank Nagy =Fermilab Research Division EED/Controls =FNAL::NAGY.HEPNET or NAGY@FNAL.Bitnet ------------------------------ Date: 20 May 87 09:09:00 EDT From: "FALCON::NIELAND" <nieland%falcon.decnet@wpafb-aamrl.arpa> Reply-to: "FALCON::NIELAND" <nieland%falcon.decnet@wpafb-aamrl.arpa> Subject: Request for VMS FINGER Program Does anyone out there have a Unix-like FINGER program for VAX VMS Version 4? I would prefer one that is written in either MACRO or FORTRAN as that is all I currently have available to me. If the program is not too long please mail it to me at TNIELAND@WPAFB-AAMRL.ARPA. Ted Nieland Systems Research Laboartories, Inc. Dayton, OH 45440 (513) 255-5156 TNIELAND@WPAFB-AAMRL.ARPA NIELAND%FALCON.DECNET@WPAFB-AAMRL.ARPA ------------------------------ Date: 20 May 87 08:25:00 PDT From: "Oberman, Kevin" <oberman@lll-icdc.arpa> Reply-to: "Oberman, Kevin" <oberman@lll-icdc.arpa> Subject: BIG backup tapes >I hear rumours of new tape drives offering >2000 MBytes of storage, using video >tapes or cartridges. Does anyone have hard information about such things, >and in particular does anyone have practical experience in using them? The new tape system is from Honeywell Test Instruments Division. The only info I have on the subject is from the May 4, 1987 Digital Review, pg. 17. It's clear that a turnkey system is not ready yet, but when it is it should be great. If it works as advertized, I sure want one. Cost: .21 cents/MByte 5.2GByte capacity 4MByte/sec transfer rate Uses VHS T-120 tape (High grade) at about $7/tape. Bit error rate of 1 in 10**12. It clearly does not use a `standard' videotape drive. The article said the a T-120 tape is written in 20 minutes. I sent off to Honeywell for more info, but I haven't received any yet. (I just sent in the request last week.) R. Kevin Oberman Lawrence Livermore National Laboratory arpa: oberman@lll-icdc.arpa (415) 422-6955 Disclaimer: I an not endorsing any product. I work for Uncle Sam and to do so would make him upset. I further take no resposibility for the accuracy of this information. I believe it is correct, but if it's not I can only say "Sorry". ------------------------------ Date: 20 May 87 13:58:18 GMT From: okstate!gregg@RUTGERS.EDU (Gregg Wonderly) Subject: Re: TPU virtual memory loss problem (Info-VAX 12-May-1987) There is an easy solution to this problem. It appears that TPU attempts to speed allocation and deallocation of data by waiting until procedure exit to clean house. I have included the output from 2 different runs of the posted procedure with the second having one modification. I moved the lines: ERASE(CREATE_RANGE(MARK(NONE),MARK(NONE),NONE)); COPY_TEXT(S); into a procedure by themselves. The result was that the test procedure ran to completion in just under 3 minutes, instead of 22 hours. And, there where only 647 page faults instead of 47 million. Needless to say, I am unimpressed with this type of difference in performance. Now off to look at search and replace in EVE, maybe there is a speed up there. ---------------------- First run ------------------------ Date and Time Iterations I/O CPU Page flts Ph.Mem 15-MAY-1987 11:16:39.31 1 159 0 00:00:23.81 708 655 15-MAY-1987 11:17:58.33 2 164 0 00:01:18.13 801 681 15-MAY-1987 11:19:44.63 3 166 0 00:02:48.94 897 713 15-MAY-1987 11:22:00.17 4 168 0 00:04:51.27 986 802 15-MAY-1987 11:25:03.40 5 170 0 00:07:27.85 1092 886 15-MAY-1987 11:28:47.53 6 172 0 00:10:38.18 1181 975 15-MAY-1987 11:33:08.68 7 174 0 00:14:24.07 1277 1064 15-MAY-1987 11:38:38.51 8 176 0 00:18:43.41 1366 1153 15-MAY-1987 11:44:54.10 9 178 0 00:23:37.95 1467 1242 15-MAY-1987 11:51:18.62 10 180 0 00:29:05.20 1565 1333 15-MAY-1987 11:58:19.01 11 182 0 00:35:07.12 1654 1422 15-MAY-1987 12:06:13.90 12 184 0 00:41:42.55 1750 1511 15-MAY-1987 12:15:14.66 13 186 0 00:48:53.89 1886 1583 15-MAY-1987 12:27:00.47 14 188 0 00:56:38.74 1979 1607 15-MAY-1987 12:37:13.19 15 190 0 01:04:55.83 2074 1607 15-MAY-1987 12:48:32.26 16 192 0 01:13:49.03 2169 1701 15-MAY-1987 12:59:17.88 17 194 0 01:23:14.24 2257 1789 15-MAY-1987 13:10:34.46 18 196 0 01:33:13.07 2353 1879 15-MAY-1987 13:23:35.55 19 198 0 01:43:51.79 2450 1969 15-MAY-1987 13:37:19.24 20 200 0 01:55:00.61 2540 2059 15-MAY-1987 13:59:25.96 21 202 0 02:06:51.60 10496 1845 15-MAY-1987 14:18:16.82 22 204 0 02:19:15.74 18157 1915 15-MAY-1987 14:40:53.10 23 206 0 02:32:20.31 36568 2036 15-MAY-1987 15:03:25.44 24 208 0 02:46:23.64 61633 2071 15-MAY-1987 15:22:14.02 25 210 0 03:00:50.16 68910 500 15-MAY-1987 15:42:27.31 26 212 0 03:15:41.13 88509 2234 15-MAY-1987 16:08:38.95 27 214 0 03:31:49.61 188868 2321 15-MAY-1987 16:54:31.61 28 216 0 03:50:56.84 488930 2443 15-MAY-1987 17:13:43.33 29 218 0 04:07:13.87 491327 2600 15-MAY-1987 17:33:44.40 30 220 0 04:24:06.95 497713 2566 15-MAY-1987 18:02:29.98 31 222 0 04:43:05.31 654311 2600 15-MAY-1987 18:51:09.28 32 224 0 05:13:04.77 1809859 2600 15-MAY-1987 20:06:02.67 33 226 0 05:49:45.30 3555137 2600 15-MAY-1987 21:09:43.71 34 228 0 06:26:33.18 5292352 2600 15-MAY-1987 21:49:36.01 35 230 0 07:00:18.75 6735915 2600 15-MAY-1987 22:35:25.70 36 232 0 07:39:32.16 8670622 500 15-MAY-1987 23:33:06.10 37 234 0 08:29:55.22 11704326 2600 16-MAY-1987 00:18:27.39 38 236 0 09:08:10.09 13424848 2600 16-MAY-1987 01:21:06.63 39 238 0 09:53:27.92 15779785 2600 16-MAY-1987 02:20:27.18 40 240 0 10:40:09.90 18134344 2600 16-MAY-1987 03:33:17.20 41 242 0 11:35:53.09 21332083 2450 16-MAY-1987 04:49:48.76 42 244 0 12:33:32.74 24603502 500 16-MAY-1987 06:02:35.12 43 246 0 13:33:02.47 28088420 500 16-MAY-1987 07:24:34.83 44 248 0 14:33:54.77 31633838 2052 16-MAY-1987 09:13:44.16 45 250 0 15:38:53.22 35188094 2600 16-MAY-1987 12:01:10.77 46 252 0 16:52:32.98 37615322 1731 16-MAY-1987 15:20:52.03 47 254 0 18:09:44.60 39604420 1548 16-MAY-1987 19:47:04.20 48 256 0 19:32:14.77 40230435 1044 16-MAY-1987 22:12:48.63 49 258 0 20:45:51.36 43498464 2598 16-MAY-1987 23:54:43.31 50 260 0 21:58:35.06 47498545 2599 -------------------------- differences output ----------------------------- ************ File MATH$STAFF:[GREGG]TRY1.COM;2 26 ERASE(CREATE_RANGE(MARK(NONE),MARK(NONE),NONE)); 27 COPY_TEXT(S); 28 ENDLOOP; ****** File MATH$STAFF:[GREGG]TRY.COM;3 26 do_func(s); 27 ENDLOOP; ************ ************ File MATH$STAFF:[GREGG]TRY1.COM;2 32 GROW ****** File MATH$STAFF:[GREGG]TRY.COM;3 31 32 PROCEDURE do_func (s) 33 ERASE(CREATE_RANGE(MARK(NONE),MARK(NONE),NONE)); 34 COPY_TEXT(S); 35 ENDPROCEDURE; 36 37 GROW ************ Number of difference sections found: 2 Number of difference records found: 8 DIFFERENCES /IGNORE=()/MERGED=1/OUTPUT=TRY.DIFF;1- TRY1.COM;2- TRY.COM;3 ------------------------------ Second run after change --------------------- Date and Time Iterations I/O CPU Page flts Ph.Mem 18-MAY-1987 15:28:24.62 0 160 0 00:00:04.25 623 546 18-MAY-1987 15:31:16.63 1 162 0 00:00:07.02 645 568 18-MAY-1987 15:33:35.40 2 164 0 00:00:09.76 645 568 18-MAY-1987 15:37:33.33 3 166 0 00:00:12.57 645 568 18-MAY-1987 15:40:52.12 4 168 0 00:00:15.43 645 568 18-MAY-1987 15:43:34.04 5 170 0 00:00:18.28 645 568 18-MAY-1987 15:46:15.55 6 172 0 00:00:21.18 645 568 18-MAY-1987 15:49:08.48 7 174 0 00:00:24.12 645 568 18-MAY-1987 15:51:20.45 8 176 0 00:00:27.06 645 568 18-MAY-1987 15:54:18.05 9 178 0 00:00:30.05 645 568 18-MAY-1987 15:56:51.20 10 180 0 00:00:33.06 645 568 18-MAY-1987 16:01:38.81 11 182 0 00:00:36.08 645 568 18-MAY-1987 16:06:23.36 12 184 0 00:00:39.20 645 568 18-MAY-1987 16:08:51.35 13 186 0 00:00:42.29 645 568 18-MAY-1987 16:11:13.13 14 188 0 00:00:45.43 645 568 18-MAY-1987 16:13:20.41 15 190 0 00:00:48.60 645 568 18-MAY-1987 16:15:26.15 16 192 0 00:00:51.82 645 568 18-MAY-1987 16:18:02.51 17 194 0 00:00:55.11 645 568 18-MAY-1987 16:20:30.74 18 196 0 00:00:58.36 645 568 18-MAY-1987 16:22:59.25 19 198 0 00:01:01.64 645 568 18-MAY-1987 16:25:34.95 20 200 0 00:01:04.97 645 568 18-MAY-1987 16:28:46.99 21 202 0 00:01:08.36 645 568 18-MAY-1987 16:31:55.58 22 204 0 00:01:11.79 645 568 18-MAY-1987 16:34:39.66 23 206 0 00:01:15.23 645 568 18-MAY-1987 16:37:00.94 24 208 0 00:01:18.71 645 568 18-MAY-1987 16:39:45.06 25 210 0 00:01:22.25 645 568 18-MAY-1987 16:42:13.66 26 212 0 00:01:25.80 645 568 18-MAY-1987 16:44:04.34 27 214 0 00:01:29.35 645 568 18-MAY-1987 16:46:33.19 28 216 0 00:01:32.94 645 568 18-MAY-1987 16:49:01.96 29 218 0 00:01:36.54 645 568 18-MAY-1987 16:51:08.91 30 220 0 00:01:40.17 645 568 18-MAY-1987 16:54:37.04 31 222 0 00:01:43.81 645 568 18-MAY-1987 16:57:11.75 32 224 0 00:01:47.50 645 568 18-MAY-1987 17:00:09.49 33 226 0 00:01:51.26 645 568 18-MAY-1987 17:02:45.31 34 228 0 00:01:55.03 645 568 18-MAY-1987 17:05:06.61 35 230 0 00:01:58.81 645 568 18-MAY-1987 17:07:20.04 36 232 0 00:02:02.67 645 568 18-MAY-1987 17:09:55.10 37 234 0 00:02:06.51 645 568 18-MAY-1987 17:12:39.15 38 236 0 00:02:10.37 645 568 18-MAY-1987 17:15:08.99 39 238 0 00:02:14.31 645 568 18-MAY-1987 17:17:39.07 40 240 0 00:02:18.24 645 568 18-MAY-1987 17:20:14.21 41 242 0 00:02:22.26 646 569 18-MAY-1987 17:23:26.19 42 244 0 00:02:26.31 646 569 18-MAY-1987 17:26:09.78 43 246 0 00:02:30.36 646 569 18-MAY-1987 17:28:30.65 44 248 0 00:02:34.48 646 569 18-MAY-1987 17:30:50.96 45 250 0 00:02:38.60 647 570 18-MAY-1987 17:33:28.37 46 252 0 00:02:42.72 647 570 18-MAY-1987 17:36:00.95 47 254 0 00:02:46.90 647 570 18-MAY-1987 17:39:45.49 48 256 0 00:02:51.08 647 570 18-MAY-1987 17:42:51.45 49 258 0 00:02:55.33 647 570 18-MAY-1987 17:46:10.70 50 260 0 00:02:59.64 647 570 ------------------------------ End of Info-Vax Digest **********************