innovus@maccs.UUCP (Anova Corp.) (02/10/88)
HP9000 series 500 (HP-UX) users:
HP has just announced the "discontinuance" of the 500 series, and
that under the standard N year support on hardware, the o/s will be frozen at
HP-UX Rev 5.2 (SYS V release 2).
I can quite understand HP placing emphasis on the 3xx and in particular the
new 8xx systems, but for those 5xx users out there I would pose the following
questions:
1.) What is the impact of being frozen at 5.2 rather than
at say the equivalent of U**x(TM?) V release 3 with
such niceties as shared libraries and streams??
2.) Will you upgrade to the 8xx series?
3.) What hardships will this place on you if any?
I am attempting here to generate a response from the user community,
and invite comments from all. I for one will miss the little beastie.
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
not a flame ... just a glimmer ... {the attitudes expressed are no-one's
R. Sciuk - INNOVUS. in particular ...}zaphod@deepthot.UUCP (millions of atoms of Lance) (02/10/88)
In article <987@maccs.UUCP> innovus@maccs.UUCP (INNOVUS Inc.) writes: > ... under the standard N year support on hardware, the o/s will be frozen at >HP-UX Rev 5.2 (SYS V release 2). > >I can quite understand HP placing emphasis on the 3xx and in particular the >new 8xx systems, but for those 5xx users out there I would pose the following >questions: > > 1.) What is the impact of being frozen at 5.2 rather than > at say the equivalent of U**x(TM?) V release 3 with > such niceties as shared libraries and streams?? Let's be honest here, 5.2 was bent, BADLY. In fact 5.2 was so buggy that our software rep. told us to NOT use it and remain at 5.1. Unfortunatly he told us that 4 months after we had installed so I have hands on experience with 5.2's bugs. So basically, HP's decision to freeze at 5.2 is in reality a decision to freeze at 5.1. > > 2.) Will you upgrade to the 8xx series? since a lot of the s/w we use is yet on the 800, no > > 3.) What hardships will this place on you if any? I suggest that HP either patch 5.2 or freeze at 5.3 For them to release 5.2, then tell us to NOT use it, and then tell us that 5.2 is the version they will be releasing at is silly. disclaimer.h #undef Robart'sResponsibleForMyThoughts #undef FlameAtHP #def AreYouListeningHP? -- humbly yours, Lance Bailey Univ. Western Ontario | Robarts Research Institute Dept. of Computer Science | Clinical Trials Unit Graduate Studies | PO Box 5015 London, Canada | London, Canada N6A 5B7 | N6A 5K8 decvax!{utcs|utzoo|watmath}!deepthot!zaphod -or- zaphod@deepthot.uucp
bruce@bnr-vpa.UUCP (Bruce Townsend) (02/11/88)
In article <1018@deepthot.UUCP> zaphod@deepthot.UUCP writes: > Let's be honest here, 5.2 was bent, BADLY. In fact 5.2 was so buggy > that our software rep. told us to NOT use it and remain at 5.1. > Unfortunatly he told us that 4 months after we had installed so > I have hands on experience with 5.2's bugs. We have upgraded to 5.2 (HP9000 s500), and have not noticed any bugs. However, we have not been using the machine for much software development. Could you please elaborate on the nature of the bugs introduced by 5.2, in order to help us unsuspecting users avoid problems? Thanks, -- Bruce Townsend (bnr-vpa!bruce) Phone: (613) 726-3008 Bell-Northern Research Usenet: {utzoo, utcs}!bnr-vpa!bruce P.O. Box 3511, Station C, Ottawa, Ontario, Canada, K1Y 4H7
mcb@hpfcls.HP.COM (Mike Berry) (02/11/88)
/ innovus@maccs.UUCP (Anova Corp.) / 5:19 pm Feb 9, 1988 / Re: HP9000 series 500 (HP-UX) users: > 1.) What is the impact of being frozen at 5.2 rather than > at say the equivalent of U**x(TM?) V release 3 with > such niceties as shared libraries and streams?? Just thought I would let you know that the s500 has had shared libraries since ~4.0. Libc and some other goodies are loaded at powerup (/usr/lib/sslibs/~). Provisions were never made for users in this regard due to the potential support problems. Mike Berry, HP, Ft Collins, CO
zaphod@deepthot.UUCP (millions of atoms of Lance) (02/12/88)
>In article <1018@deepthot.UUCP> zaphod@deepthot.UUCP writes: >> I have hands on experience with 5.2's bugs. > In article <191@bnr-vpa.UUCP> bruce@bnr-vpa.UUCP (Bruce Townsend) writes: > ... Could you please elaborate on the nature of >the bugs introduced by 5.2 ... apparently a new "has the file system been checked since last change" flag was added to the super-block (i don't have the HP reference to this near me) and the first indication of weirdness is that our 7935 would not mount at boot time. After a fsck on the device it would mount. a lot of the /usr/contrib/bin progs would not work, such as users (mem fault and core dump) dir (core dump) catman -w (core dump) fortune (hang) adventure (hang) now, granted, the last two do not generate a major crisis, but since nearly every user i've met has fortune in their .login/.profile, they would all "get stuck" when they tried to log in. sigh. there were more problems, but i am sure that you get the idea, for a while it seemed that half the world was hanging and half the world core dumping. this is why, as i stated in my first posting, that our software rep. advised all his customers to NOT use 5.2, and thus why I feel that in all respects, HP has frozen at 5.1 not 5.2. -- humbly yours, Lance Bailey Univ. Western Ontario | Robarts Research Institute Dept. of Computer Science | Clinical Trials Unit Graduate Studies | PO Box 5015 London, Canada | London, Canada N6A 5B7 | N6A 5K8 decvax!{utcs|utzoo|watmath}!deepthot!zaphod -or- zaphod@deepthot.uucp
steve@polyslo.UUCP (Steve DeJarnett) (02/15/88)
In article <191@bnr-vpa.UUCP> bruce@bnr-vpa.UUCP (Bruce Townsend) writes: >In article <1018@deepthot.UUCP> zaphod@deepthot.UUCP writes: >> Let's be honest here, 5.2 was bent, BADLY. In fact 5.2 was so buggy >> that our software rep. told us to NOT use it and remain at 5.1. >> Unfortunatly he told us that 4 months after we had installed so >> I have hands on experience with 5.2's bugs. > Could you please elaborate on the nature of > the bugs introduced by 5.2, in order to help us unsuspecting users > avoid problems? Yes, please. I'm going to install 5.2 (I think, unless I hear more bad news from the net) in about 3 weeks (quarter break). If you know of any problems, I'd really like to hear about it now, while I can still decide to hold off on installing. I really don't want a line of irate users standing outside my office next quarter wanting to know why their programs don't work/ aren't there/whatever. Please, let us benefit from you troubles. Thanks in advance, ------------------------------------------------------------------------------- | Steve DeJarnett | ...!ihnp4!csun!polyslo!steve | | Computer Systems Lab | ...!{csustan,csun,sdsu}!polyslo!steve | | Cal Poly State Univ. | ...!ucbvax!voder!polyslo!steve | | San Luis Obispo, CA 93407 | | ------------------------------------------------------------------------------- #include <std_disclaimer.h>
web@hpubmaa.HP.COM (Bill Bennett) (02/19/88)
> Yes, please. I'm going to install 5.2 (I think, unless I hear more >bad news from the net) in about 3 weeks (quarter break). If you know of any >problems, I'd really like to hear about it now, while I can still decide to >hold off on installing. > Steve DeJarnett There were three main reasons for the 5.2 update: 1. To provide support for the Ada compiler 2. To correct known bugs 3. To provide command parity with the Series 300 & 800 Since several problems were found, the official recommendation was to NOT update, unless you needed Ada, or some of the other bug fixes. >I really don't want a line of irate users standing outside my office next >quarter wanting to know why their programs don't work/aren't there/whatever User programs should not be affected. (He says 8-) As far as fortune/adventure and others go, I think they were using shared libraries. Because of the way this is implemented, any program using a shared library will need to be relinked. Since these programs are not part of HP-UX, new versions were not shipped with 5.2. I'd suggest you call your local HP support person, and get a copy of the Update_info document. This describes all of the changes in 5.2. After reading it, you can decide whether you wish to do the update. Bill Bennett hpfcse!hpubmaa!web
rclark@bgphp1.UUCP (Roger N. Clark) (02/20/88)
> HP9000 series 500 (HP-UX) users: > HP has just announced the "discontinuance" of the 500 series, and > that under the standard N year support on hardware, the o/s will be frozen at > HP-UX Rev 5.2 (SYS V release 2). I am QUITE UPSET!!!! I bought a 500 in late 1984. The machine was only introduced a couple of years earlier. It was becomming obvious by 1986 that the 500 was on its way out. In 1987 it was certain. Now it is here. First I do not know anyone who will buy a new computer until it "matures" for a year or two. If HP's computers only last 6 years from intruduction, and you wait a year or two before buying, then you are stuck a few years later. Contrast this to other computer manufacturers. DEC still supports PDP 11's, and is still comming out with new products for it! Gould still supports their old SEL's; Concurrent (Perkin Elmer) still supports their PE3200's. From what I have heard from HP people, the ONLY reason that HP sold the 500 for so long is that the NAVY was buying a lot of them. Now the question is whether to upgrade to an 800. How do we know if HP will continue the 800 or will it be phased out in 3 or 4 years? HP will probably say no way, but there was no indication of that when we bought the 500s. I think HP's reputation will be damaged badly in the eyes of their current customers. HP seems to have a good products, even in the 500. The prices are competitive, and the maintenance costs are great. But if you have to replace your machine every 3 or 4 years, it is just not worth it. It is about 4 years from the introduction that HP stopped porting software to the 500. You can't get NFS or TCP/IP (Except for an outrageous price you can buy TCP/IP from Wolongong) unless you upgrade to the 800 or 300. How do we get domain based mail for the 500? I had a quote done for an upgrade from my 500 to an 800. HP gave an upgrade trade in of $9,196 for the 500, bringing the 800 price (with the boards I need to have the equivalent of what I have now) to $40,000. That hardware retails for about $60,000! A large fraction of the upgrade price is for software! The upgrades are so poor there is little incentive. What is the NAVY going to do? So I can't decide whether to upgrade to an 825, buy 350's (including a few diskless workstations) or forget HP and try for another vendor (of course any purchase would be by competition). Many of the people I work with are buying Suns. Porting between HP and Sun is not always easy. I have a strong incentive for going with Sun. I think if HP wants to keep its customers and a good reputation, they will come up with a better deal. Roger N. Clark (Any opinions expressed here are mine and not necessarily those of the USGS) {known-world}!hplabs!hpfcla!hpfcse!hpuecoa!bgphp1!rclark
ckw@hpcupt1.HP.COM (Chick Webb) (02/26/88)
>> HP has just announced the "discontinuance" of the 500 series, and >> that under the standard N year support on hardware, the o/s will be frozen >> at HP-UX Rev 5.2 (SYS V release 2). >Contrast this to other computer manufacturers. DEC still supports >PDP 11's, and is still comming out with new products for it! Gould >still supports their old SEL's; Concurrent (Perkin Elmer) still >supports their PE3200's. HP has *not* stopped supporting the S500, we've just stopped *selling* them. As with all of our computing products, we remain committed to *supporting* them for 5 years following obsolesence. >Now the question is whether to upgrade to an 800. How do we know if >HP will continue the 800 or will it be phased out in 3 or 4 years? >HP will probably say no way, but there was no indication of that >when we bought the 500s. I think HP's reputation will be damaged >badly in the eyes of their current customers. HP's committment to the S800 product line should be obvious by now. In the span of barely two years we have introduced three *distinct* members of this family, with performance ranging from 3-7 (VAX) MIPS. (Contrast this to the S500, which was a single point product.) You can bet that more offerings, on both ends of the performance scale, are in the works. As our venerable CEO has stated publicly, the HPPA Architecture is the future of HP computing. >HP seems to have a good products, even in the 500. The prices are >competitive, and the maintenance costs are great. But if you have >to replace your machine every 3 or 4 years, it is just not worth it. You don't *have* to replace it! In fact, I'd be willing to bet that that thing will be running long after you *want* (or *need*) to replace it! >It is about 4 years from the introduction that HP stopped porting >software to the 500. You can't get NFS or TCP/IP (Except for an >outrageous price you can buy TCP/IP from Wolongong) unless you >upgrade to the 800 or 300. How do we get domain based mail for the >500? The S500 was first introduce in 1982. This is 1988. That's more like 6 years by my (HP) calculator. The issue of bundled TCP/IP was a much debated subject, and certainly the end result is one that everyone doesn't agree with. :-( >I had a quote done for an upgrade from my 500 to an 800. HP gave an >upgrade trade in of $9,196 for the 500, bringing the 800 price (with >the boards I need to have the equivalent of what I have now) to >$40,000. That hardware retails for about $60,000! >A large fraction of the upgrade price is for software! >The upgrades are so poor there is little incentive. Why do you say this? For your money you'll be getting: - Faster hardware - More memory - Better reliability - Bundled networking (saves you $10K on Wolongong) - Source-code compatiblity with future products - much, much more! If you look at our price/performance, and especially cost of ownership, I think you'll find that HP is still a good buy! As for most of the cost being for software, that is debatable, but even if it weren't, do you think that software grows on trees? >So I can't decide whether to upgrade to an 825, buy 350's (including >a few diskless workstations) or forget HP and try for another vendor >(of course any purchase would be by competition). Many of the >people I work with are buying Suns. Porting between HP and Sun is >not always easy. I have a strong incentive for going with Sun. Weigh your options carefully, especially in the area of cost of ownership. >Roger N. Clark ---------- Chick Webb HP-UX Technical Support Hewlett-Packard Company {ucbvax, etc.}!hpda!ckw P.S. Support my bad habits, buy HP!
mcb@hpfcls.HP.COM (Mike Berry) (02/27/88)
Re: Support life of s500 ckw@hpcupt1.HP.COM (Chick Webb) says: > HP has *not* stopped supporting the S500, we've just stopped *selling* > them. As with all of our computing products, we remain committed > to *supporting* them for 5 years following obsolesence. I believe our support life is 10 years. My efforts to obtain support lives for our competition failed. This quantity is something to consider in any vendor that you purchase from. Mike Berry, Ft Collins HP
irf@kuling.UUCP (Bo Thide) (03/08/88)
When we bought our 9000/540 little over 4 years ago we heard about
the new FOCUS II CPU, so we started out with a single FOCUS I. Later
when we upgraded to include also a FOCUS II plus 2 more MB of ("commercial")
RAM we experienced a tremendous increase in throughput. What we then hoped
for was of course a FOCUS III (with a higher clock frequency etc), but no,
instead HP discontinued the 500 series altogether....
We think that was an unwise decision, especially since the 500 is something
of an engineering masterpiece. The finstrate cards are really something to
impress hardware people with ... Not only do they look nice but they run
and run. In fact, in all these years we haven't had any hardware problems
whatsoever with our 540. And the multi-CPU configuration turned it into a real
workhorse being able to support a large number of users (typically 10-15)
simultaneously without any noticable degradation in turn-around.
What's annoying about the 500 series is that the software is not being kept
up-to-date, no TCP/IP, no X Windows, no troff (!!), no (supported) tplot,
plot, graph etc, and the hardware (CPU, RAM, interfaces) is VERY expensive
-- somehow we have a feeling prices have gone up rather than down...
Our new 350 is much faster but I doubt that it is a good replacement as
a central computer for our whole institute. We don't even think an 800
series can compete with a 500 with 2 or 3 computers in terms of multi-user
capablility. So the logical question here is: Why doesn't HP have a
multi-CPU 800? Was the experience with the 500 series so bad that HP has
given up that idea?
--
>>> Bo Thide', Swedish Institute of Space Physics, S-755 90 Uppsala, Sweden <<< Phone (+46) 18-300020. Telex: 76036 (IRFUPP S). UUCP: ..enea!kuling!irfu!btzaphod@deepthot.UUCP (millions of atoms of Lance) (03/14/88)
In article <672@kuling.UUCP> irf@kuling.UUCP (Bo Thide) writes: > ... We don't even think an 800 >series can compete with a 500 with 2 or 3 computers in terms of multi-user >capablility. So the logical question here is: Why doesn't HP have a >multi-CPU 800? Was the experience with the 500 series so bad that HP has >given up that idea? I agree. Our main application of our 2 CPU 9000/550 is data collection, at 1am a schedular starts up that phones 50 PC's around North America and exchanges data with them. The only action between the two machines is the simultaneous of information (grouped together as a single file). To save time we have multiple modems, allowing concurrent sessions. when we went from one modem/session-at-a-time to two, we noticed a definite throughput increase (sessions acheived/ time period) . when we went from two to three, we did not not an improvement, but rather a decrease in thoughput as communication sessions began to time out while two CPU tried to do three i-o intensive operations. to acheive even better throughput, we are tacking in a third CPU. on applications such as ours (concurrent near-identical sessions), a multi CPU machine is a neccessity. -- humbly yours, Lance Bailey UWO, Dept. of Comp. Science | Robarts Research Institute Graduate Studies | Clinical Trials Resources Group London, Canada | PO Box 5015, 100 Perth Dr. N6A 5B7 | London, Canada, N6A 5K8 decvax!{utcs|utzoo|watmath}!deepthot!zaphod -or- zaphod@deepthot.uucp
dc@gcm (Dave Caswell) (03/16/88)
In article <672@kuling.UUCP> irf@kuling.UUCP (Bo Thide) writes:
:When we bought our 9000/540 little over 4 years ago we heard about
:the new FOCUS II CPU, so we started out with a single FOCUS I. Later
:when we upgraded to include also a FOCUS II plus 2 more MB of ("commercial")
:RAM we experienced a tremendous increase in throughput. What we then hoped
:for was of course a FOCUS III (with a higher clock frequency etc), but no,
:instead HP discontinued the 500 series altogether....
:
:We think that was an unwise decision, especially since the 500 is something
:of an engineering masterpiece. The finstrate cards are really something to
:impress hardware people with ... Not only do they look nice but they run
:and run. In fact, in all these years we haven't had any hardware problems
:whatsoever with our 540. And the multi-CPU configuration turned it into a real
:workhorse being able to support a large number of users (typically 10-15)
:simultaneously without any noticable degradation in turn-around.
Let your users try a SUN 3 for a day and see if they feel the same way
about performance. Try the basic benchmark. How many times can the
date command run in a second. 12 on a Sun 3 on a HP (using c-shell on both).
Time to compile the empty program main() {} over twice as long on the HP etc.
I'm sure that just like any other vendor people's experiences vary widely.
Below is a partial list of problems I encountered in at two different
companies using HP-UX on a series 500.
I have walked into the computer room and been able to see and smell the smoke
coming from those gold plated circuit boards. At another company we were
plagued with frequent system crashes, easily over 20 in a one year period.
There is nothing like having all your work end as the system console is
filled with rows and rows of digits; when HPUX crashes a "tombstone" (a
list of values in memory) is produced.
I have a four line C program and yes three of them are main { } that when
compiled with the right options will crash the system. This is unrelated
to the crashing problem above.
The C shell besides being incredibly slow would often go into an infinite
loop and have to be killed.
Curses was so buggy as to be unusable. It wouldn't keep the screen
consistent.
FILE system problems.
Two files of the same name in a directory. Type "ls a" and see two files
listed. "Mv a b" and then "mv a c" fixed the problem.
Fsck saying "remove inode with size zero?" and then immediately exiting
without giving the user a chance to answer.
Fsck saying something like "invalid file / remove?". I didn't have the
nerve to answer that one at all for a while.
Rm saying something like "can not delete non-existent file".
You can delete it by using fsdb to change its inode type to 5 and
then doing a fsck to remove an unallocated inode. "ls -l" first prints
the non-existent files and then the existing files.
Once the creation date of every file in a directory was changed to the same
date (and no we didn't say touch *).
Once the disk got completely corrupted with hundreds of files missing or
liked into lost and found.
The strange thing is that all these file system problems happened during
"normal use". We have had at least five power failures and things like
that without even a problem. I would be just going through the edit,
comiles and debug route and do an ls of my directory and see a file listed
twice. Naturally your first thought is that it has an embedded unprintable
character, but no.jim@cs.strath.ac.uk (Jim Reid) (03/16/88)
In article <672@kuling.UUCP> irf@kuling.UUCP (Bo Thide) writes: >............. So the logical question here is: Why doesn't HP have a >multi-CPU 800? Was the experience with the 500 series so bad that HP has >given up that idea? No, it's probably because building closely-coupled multiprocessors that use RISC (eg the 800 series CPU) isn't easy. RISC architectures like to have on-chip caches, N-deep pipelines and all sorts of things like that. This causes horrendous problems when you try to keep the value of shared variables (eg kernel data structures) consistent. eg CPU #X updates some variable, while CPU #Y still has an old copy in its cache and CPU #Z is just about to fetch an even older copy from its register file. Jim -- ARPA: jim%cs.strath.ac.uk@ucl-cs.arpa, jim@cs.strath.ac.uk UUCP: jim@strath-cs.uucp, ...!uunet!mcvax!ukc!strath-cs!jim JANET: jim@uk.ac.strath.cs "JANET domain ordering is swapped around so's there'd be some use for rev(1)!"
campbelr@hpsel1.HP.COM (Bob Campbell) (03/17/88)
While HP does not currently have a multiprocessor 9000s800, it is by no
means ruled out. From the "Precision Architecture and Instruction
Reference Manual" (Part # 09740-90014)
The processor hardware is optimized for the simple, often used
instructions that execute in a single CPU cycle. Implementation
of more complex functions is assigned to system software or to
processor assists such as the floating-point coprocessor. Multi-
processors are also supported and may be used for higher performance
or fault-tolerant computing.
If you know where you can see a 9000s850 you can see that the idea has been
thought of. If you would really like to see MP support, tell your sales
rep to pass your word along.
Bob Campbell Some times I wish that I could stop you from
Hewlett Packard talking, when I hear the silly things you say.
campbelr@hpda.HP.COM - Elvis Costellocongdon@gargoyle.UChicago.EDU (Richard Congdon) (03/17/88)
Before my comments, let me say that we have a 540 and an 825S, and that I have found a world of difference... >... The finstrate cards are really something to >impress hardware people with ... Not only do they look nice but they run >and run. In fact, in all these years we haven't had any hardware problems >whatsoever with our 540. And the multi-CPU configuration turned it into a real >workhorse being able to support a large number of users (typically 10-15) >simultaneously without any noticable degradation in turn-around. Yes the cards look nice; especially the old 256K cards. I also have had no hardware problems, or a single crash! >What's annoying about the 500 series is that the software is not being kept >up-to-date, no TCP/IP, no X Windows, no troff (!!), no (supported) tplot, >plot, graph etc, and the hardware (CPU, RAM, interfaces) is VERY expensive >-- somehow we have a feeling prices have gone up rather than down... In the past, HP has been rather bad about making their machines work with other vendors hardware/software. (This seems to be changing, though.) Most of the software you mentioned is available in various ways (not X Windows). I have never understood the lack of troff. I think that the hardware costs reflect their high quality, and no, they don't come down unless they are selling it. Also, once they come out with something better, any hope of price reductions is lost. >... We don't even think an 800 >series can compete with a 500 with 2 or 3 computers in terms of multi-user >capablility. On this point, I will have to strongly disagree. I have found our 825S to be a much faster machine than the 540. On non-IO intensive things, it is about 9-10 times faster than the 540; IO-intensive is about 3 times faster. It is much, much better on handling 10-15 users (our peak crowd). Your experience may be different, but I found that the 540 really bogged down with over 10 users (unless they were all using vi). -- Richard Congdon Dept. of Education, Univ. of Chicago ...ihnp4!gargoyle!paideia!{richard,root} (312) 702-9453
neil@yc.estec.nl (Neil Dixon) (03/17/88)
In article <672@kuling.UUCP> irf@kuling.UUCP (Bo Thide) writes: >-- somehow we have a feeling prices have gone up rather than down... > >Our new 350 is much faster but I doubt that it is a good replacement as >a central computer for our whole institute. We don't even think an 800 >series can compete with a 500 with 2 or 3 computers in terms of multi-user >capablility. So the logical question here is: Why doesn't HP have a >multi-CPU 800? Was the experience with the 500 series so bad that HP has >given up that idea? >-- Intrigued by this, I ran the MUSBUS benchmarks on our 520, which has 3 cpu's and *Mb of memory, and compared the results with our 825. Below are our figures for the work load test:- 520 825 1 Concurrent User Elapsed time 377.10 Elapsed time 370.67 CPU time 88.83 CPU time 12.60 4 Concurrent Users Elapsed time 397.10 Elapsed time 372.77 CPU time 461.53 CPU time 49.47 8 Concurrent Users Elapsed time 534.27 Elapsed time 383.43 CPU time 1013.87 CPU time 99.90 I think these results speak for themselves. -- Neil Dixon <neil@yc.estec.nl> UUCP:...!mcvax!esatst!neil, BITNET: NDIXON@ESTEC Thermal Control & Life Support Division (YC) European Space Research and Technology Centre (ESTEC), Noordwijk, The Netherlands.
jules@zen.UUCP (Julian Perry) (03/19/88)
[irf@kuling.UUCP (Bo Thide):] >What's annoying about the 500 series is that the software is not being kept >up-to-date, no TCP/IP, no X Windows, no troff (!!), no (supported) tplot, >plot, graph etc, and the hardware (CPU, RAM, interfaces) is VERY expensive We agree, it's a shame to throw out 500's just because they're not the latest hardware; they were good in their day. >Our new 350 is much faster but I doubt that it is a good replacement as >a central computer for our whole institute. We don't even think an 800 >series can compete with a 500 with 2 or 3 computers in terms of multi-user >capablility. Was the experience with the 500 series so bad that HP has >given up that idea? Yes we think the 350 is great (ours is SRX-ed) and fast but we wouldn't really recommend 10-15 users...single user is much better. :-) Let us put you right on this matter of an 800 not coping with 10-15 users, the 800 series is VERY fast...we had a network of 3 series 500's (total of 6 processors (mainly Focus II's), 12MB RAM, 4 discs, 30ish users) and we thought they were fast till we replaced the lot with a 16MB 840. We always got complaints about the interactive performance of the 500s once more than 10-15 users got going, but now other users' activity is virtually unnoticeable. For example, a single TeX or LaTeX nearly stops a 500 but I ran 39 simultaneous LaTeX's (text~300K, data~600K) on the 840 and it hardly slowed down at all! [zaphod@deepthot.UUCP (millions of atoms of Lance):] >Our main application of our 2 CPU 9000/550 is data collection, at >1am a schedular starts up that phones 50 PC's around North America >and exchanges data with them. The only action between the two machines >is the simultaneous of information (grouped together as a single file). >To save time we have multiple modems, allowing concurrent sessions. >When we went from one modem/session-at-a-time to two, we noticed a >definite throughput increase (sessions acheived/ time period) . >when we went from two to three, we did not not an improvement, but >rather a decrease in thoughput as communication sessions began to >time out while two CPU tried to do three i-o intensive operations. >[...] >to acheive even better throughput, we are tacking in a third CPU. >[...] >on applications such as ours (concurrent near-identical sessions), >a multi CPU machine is a neccessity. Strictly speaking you would need multiple CPU's to get the best performance with multiple processes, but in reality a single fast CPU is sufficient.....the series 800 is FAST. We have found NO tasks at which the 500's are faster. The 840 version of HP-UX has features to make it better than the 500 version when running multiple processes -- kernel pre-emption, real-time extensions and BSD fast file system, etc.. [jim@cs.strath.ac.uk (Jim Reid):] >In article <672@kuling.UUCP> irf@kuling.UUCP (Bo Thide) writes: >>............. So the logical question here is: Why doesn't HP have a >>multi-CPU 800? Was the experience with the 500 series so bad that HP has >>given up that idea? > >No, it's probably because building closely-coupled multiprocessors that >use RISC (eg the 800 series CPU) isn't easy. Just because it isn't easy doesn't mean it can't be done. HP Precision Architecture allows assist hardware to enhance system performance; for example, current 800 series use as a co-processor the floating-point chips found on the FOCUS II board (according to HP Journal). Additionally, multi-processing is supported in HP-PA "to provide incremental performance via distribution of the system workload over multiple CPUs, or can be configured redundantly to provide fault-tolerance". "Software is still responsible for maintaining consistency for I/O, for modifying instructions and for virtual address mapping." (HP Precision Architecture Reference Manual) After all, look how smootly HP made the multi-CPU 500s perform. :-) Jules and Frank [the spectral defenders of the faith] -- IN-REAL-LIFE: Julian Perry E-MAIL: jules@zen.co.uk || ...!mcvax!ukc!zen.co.uk!jules PHONE: +44 532 489048 ext 217 ADDRESS: Zengrange Limited, Greenfield Road, Leeds, England, LS9 8DB
irf@kuling.UUCP (Bo Thide) (03/20/88)
In article <969@gargoyle.UChicago.EDU> congdon@gargoyle.uchicago.edu.UUCP (Richard Congdon) writes: > >Before my comments, let me say that we have a 540 and an 825S, and that >I have found a world of difference... > >hardware problems, or a single crash! > > . I have found our 825S to >be a much faster machine than the 540. On non-IO intensive things, it is >about 9-10 times faster than the 540; IO-intensive is about 3 times faster. >It is much, much better on handling 10-15 users (our peak crowd). Your >experience may be different, but I found that the 540 really bogged down with >over 10 users (unless they were all using vi). Well, my first posting may have given a too rosy picture about the 2 CPU 500 vs. the 1 CPU version we started with. But, there was a tremendous improvement with 10-15 people logged in. With our new 3CPU configuration we hope we have a machine that we can use until HP comes out with something very powerful that we can use as an instititue central computer. Until then we invest in 350s (/370s?), Woolongong TCP/IP (!) and, maybe a 835 for dedicated purposes. My "own" 350 with HPs FPA plus the Ariel FFT-card is also VERY fast and I use it both in my office and in field experiments. I hope the announced new 68030 upgrade card will not be too expensive since I need all the power I can get in time-critical applications. Unfortunately, I don't get all the money I want and HP only gives away computers for free to Computer Science Depts ... -Bo -- >>> Bo Thide', Swedish Institute of Space Physics, S-755 90 Uppsala, Sweden <<< Phone (+46) 18-300020. Telex: 76036 (IRFUPP S). UUCP: ..enea!kuling!irfu!bt