Re: Sparc 2 or 4/490 - SUMMARY (long)

From: Gary Blumenstein (garyb%gallium@uunet.uu.net)
Date: Tue Jun 04 1991 - 13:40:00 CDT


Dear Net Folks,

Back in April, I posted my inquiry about which is better; to buy multiple
Sparcstation 2's versus a single 4/490. The responses were extremely
useful to us, once again proving the value of Usenet and the whole Internet
community in general. I'd like to thank everyone who took the time to
write. Hopefully, some of you may find this information valuable in your
own decsision making process.

In particular, the replies which give information on configuring NFS
servers for optimum performance were the most interesting to us and
the advice is well taken here at good `ol UPS!

Finally, we have not yet reached a decision on which platform to buy. We
*are* going to request a non-disclosure meeting with Sun to discuss their
plans. This sentiment was either implied or suggested in more than one
response. In the meantime... Enjoy!

- Gary

-- 
Gary M. Blumenstein  //  UNIX Sys Admin
United Parcel Service, Research & Development
51-53 Kenosia Road Danbury, CT  06810  //  Email:  garyb@casey.UUCP

---------------------- Summary begins here ---------------------------

From: opal!uunet!archone.tamu.edu!byron (Byron Rakitzis) Organization: College of Architecture, Texas A&M University.

I administer a bunch of workstations (sgi, next, sun3 and sun4) served by a sun 4/280. If I were faced with your choice, I would *probably* go with the SS2's, simply because of the iron alligator theory you presented. Unless you have a very good reason for needing a VME machine with all the bells and whistles, I'd go for the cheaper boxes with SCSI disks.

But that's just my biased opinion.

(Also, for powerful multiprocessors I'd stay far away from sparc machines and get something reasonable like an sgi, but both of these are just pies in the sky for me, so let me dream on)

--

From: opal!uunet!curie.ces.cwru.edu!glenn (Glenn Crocker)

We are faced with the decision on whether to buy a Sun 4/490 server or get two or three Sparcstation 2's with multiple 1GB SCSI disks.

Excellent. Please post whatever replies you can. I'm looking at the Sparc 2 now, and am interested in any info you get.

I don't have any great insight for you, but I think splitting the load onto multiple Sparc 2's is better than a centralized server. With a good backup system, and automounter, your uptime should be higher overall. (i.e. when the 4/490 goes down, how much of your operation will be forced to stop? With Sparc 2's with local disks, you can most likely hack a temporary solution to having one machine going down.)

--

From: "Ric Anderson" <arizona!ric@uunet>

You did a good job of summing up the differences. For an NFS server the IPI performance doesn't buy you beans over a good SCSI disk, because the IPI's speed is all but lost in the NFS overhead.

The *MOST* important thing you can do is put root, swap, and /usr on a disk local to each workstation. SunOS is terribly swap happy, and a fast local disk (Quantum Prodrive-105 or 210 is what Sun seems to ship, and they work well) makes a major difference in net traffic and in workstation performance. The last set of stats I saw on the net showed that, for a "typical" Sun workstation, 81% of the file traffic was either running programs in /usr/bin, or swapping.

The next most important thing is keep heavily accessed files on the machines using them, so things come off of a local disk, instead of via NFS. At 40 milliseconds per 8192 bytes, NFS is a real dog for big file xfers.

The IPI based disks are worth the price of admission, if you need to transfer large blocks of data from the disk to a program running on the machine driving the disk. They key here is *LARGE* blocks of data. Most programs don't read enough bits in any given read to utilize the IPI's speed. So far, only big image files seem to show any performance advantage for the IPI disks around here.

We replaced a VAX 8650 with a Sparc-2 (64MB ram, 200MB Sun internal HD, and an S-Bus SCSI card so we would have two contollers; We added 4 Fuji 2266SA disks, and an Exabyte from third party sources). We run 30-50 users with no problem on that machine.

Our 4/490 is a file server to 30 dataless Sun Clients (mixture of Sparc-1, Sparc-1+, and Sparc-IPC), as well as a login host for the secretarial staff. It swaps itself silly (32MB of memory is *NOT* enough), but we haven't got the money to solve that problem right now. In spite of the swap ratio, response is still good.

I still haven't seen any of these machines with multiple compute intensive tasks (multiple meaning more than two or three), so I don't know how well they context switch when really pushed.

The SS2 is better bang for the buck up front, and is cheaper to add peripherals to; maintenance is also cheaper. I can't see buying another 4/490, unless someone has a serious need to do heavy work on monster files, and, even then, the costs argue for an SS2 (which has a faster CPU than a 4/490, by the way).

As for expandability: everything is changing so fast, any machine you buy is a 2-3 year throwaway. And prices are dropping, so buying a 4/490, on the off chance that you might need more cpu power, and Sun might do multiple CPUs is not attractive from where I sit (you may have a better view though).

If you need multiple CPUs today, Solborne sells such a beast (Sun CPU, and SunOS). I've never owned one, so I don't know beans about them.

Hope some of this rambling helps,

--

From: Gijs Mos <opal!uunet!bio.vu.nl!gjm> Organization: VU, Biology, The Netherlands

>We are faced with the decision on whether to buy a Sun 4/490 server or >get two or three Sparcstation 2's with multiple 1GB SCSI disks. ..

>Basically, I was hoping to get some feedback on which configuration would >make the most sense to you and why. >

There is an interesting paper on server performance available from sun in the Netherlands. Try postmaster@sun.nl . If this fails mail me at gjm@tes.nl .

Basically they say that you should use a Sparcstation II or a I+ as "read only" file server. What they mean is that you use one machine equipped with some 32 to 40 MB of memory, two slow built in disks and optionally a second ethernet interface as a server for read-only stuff e.g. executables, manuals, lib stuff and as a NIS master. The idea is simple: the frequently used parts of the OS and open-look fit in about 32 MB. After a while all of this will be in main memory. The disks will hardly be used. It is crucial that you don't use this machine for anything else. If you look at the machine the cpu will hardly be used (some 15% or so), disks will be idle so it looks like a good candidate to do some computational work. But this will immediately flush parts of the buffer cache, the disk won't be idle anymore and you'll end up with a slow server.

For stuff that gets written often (write tru for NFS!) and for stuf that is read not very often (caching won't help) they recommend a second server. For this machine a large memory won't realy help. I/O speed is crucial here. In this role a 490 equipped with IPI disks can be a good choise. That the 490 might see a second CPU in future is not important. I/O troughput is the important thing for this machine. But you should also consider the possibility of a Sparcstation II or I+ with multiple SCSI controllers.

As computational servers (again, don't combine them with your NFS servers) a Sparcstation II may be a good choise. Give it enough memory to prevent to much swapping/paging. If you use large address spaces a LOCAL disk for paging would be a good choise. Don't use the local disk for executables. They can be retrieved (paged in) from the read-only nfs server a lot faster (when in cache). I don't really now if a future multi processor machine is of interest. Seems to me that adding a second sparcstation II computational server is a better solution. Multiprocessing may help if you can use both CPU's together on the same problem. And a sparcstation II will be a lot cheaper than a second CPU and some extra memory in a 490.

>Since there are so many issues that can influence this decision, I'll start >by including a list of pros and cons that I've come up with. Your responses

>o It seems to me that multiple Sparc 2's would buy us more aggregate compute > power than a single 4/490. While the 4/490 provides greater i/o capacity, > by carefully distributing applications across two or more smaller servers, > the system could be load balanced to keep i/o activity within acceptable > performance bounds.

True, if you can split I/O over two machines (or controllers/disk) the advantage of the IPI disks is less.

>o Most of our network bandwidth is consumed by NFS traffic (largely estimated > but some benchmarks have been taken with nfswatch and nfsstat). Currently, > our NFS traffic is not very bad at all (thanks to the Automounter) IOPS on > our busiest server is in the neighborhood of 20-30 max.

This is ok. For the read-only server sun quotes up to 50 IOPS (reads!).

>o One of the big plusses for the 4/490 is the optional Prestoserve board, but > I understand Legato now makes them for the SBus. If NFS traffic increases > measurably, we could add Prestoserve's to the Sparc 2's.

A prestoserve is basically a NV memory. It makes the NFS write-tru's faster. They are ack-ed when they are in NV memory. Since writes are delayed the actual disk writes can be scheduled in a way that minimises on seek time. A prestoserve won't help for reads. So consider it for servers for files that are frequently updated. Dont use it for read-only servers. I didn't know of the Sbus variant.

>o Sparc 2's are cheaper and can always be relegated to a user's desk when they > become obsolete as servers. > >o Cost of SCSI -vs- IPI. Since we'd be paying about 3 times more for > the same number of megabytes, it seems more cost-effective to stick > with relatively cheap SCSI disks. (It seems we're *always* looking > for more space!)

Yes, and in a Sparcstation II you can add upto 3 extra SCSI controlers if you pull the framebuffer. Would make more sense to add two SCSI controllers and an extra ethernet controller.

>o We do not yet have any *real* performance monitoring statistics to warrant > the absolute need for a high-end compute server.

See the above. Mixing NFS servers and computational servers is deadly!. The crusial thing is to use as much memory as possible as data cache!

>o I am afraid the 4/490 will turn into an "iron alligator"; eating up > corporate funds to keep it happy. Most peripherals seem > [dis]proportionately more expensive when going to a centralized server.

True. However, the VME bus gives a much broader choise for the moment. .. >Here are some points that favor the 4/490. ..

>o Maintenance on the 4/490 is comparable to the maintenance we pay for > our 3/280.

So what? You should compare a 490 solytion with a Sparc II solution.

>o There is some chance that we will need an on-line database sometime in the > future (When? Nobody here knows!). It would not be for transaction > processing, but rather the need to access large, finite data sets for > engineering oriented applications. The high (6Mb/sec) transfer rate of the > IPI disks would seem better suited for this application than SCSI.

Synchronous SCSI runs a 4 Mb/sec. How large are large data sets? If large means 96 MB you're better off with SCSI and lots of memory.

>Some additional questions: > >Since we are concerned about gaining the most utility from the 4/490, what >would happen if for instance, we had set up the 4/490 as an NFS server >and later on, our users wanted to run simulations on it? Would this >platform provide enough capacity to play multiple roles? Roles that might

No. No machine would.

--

From: opal!uunet!mcnc.org!med.unc.edu!robinson (Gerard A. Robinson) Organization: UNC School of Medicine

In article <1207@gaboon.UUCP> you write: >Here are some reasons in favor of using Sparc 2's: > >o It seems to me that multiple Sparc 2's would buy us more aggregate compute > power than a single 4/490. While the 4/490 provides greater i/o capacity, > by carefully distributing applications across two or more smaller servers, > the system could be load balanced to keep i/o activity within acceptable > performance bounds.

A single SS2 will buy you more power than a 4/490. Besides, I like the redundancy. You can almost always take an application and put it up on another SS2 and slow everyone down, but let them all still work.

>Here are some points that favor the 4/490. > >o Sun sales is offering us a _substantial_ discount for trading in our old > 3/280 server. This lowers the base price of the 490 to the point where > it almost seems like the logical thing to do, if not for any other reason > than to stay current with newer technology. Also, we are afraid that > we will become stuck with the 3/280 when Sun no longer supports it.

Check with used equipment vendors and see what they'll offer for it.

>o Maintenance on the 4/490 is comparable to the maintenance we pay for > our 3/280. > >o There is some chance that we will need an on-line database sometime in the > future (When? Nobody here knows!). It would not be for transaction > processing, but rather the need to access large, finite data sets for > engineering oriented applications. The high (6Mb/sec) transfer rate of the > IPI disks would seem better suited for this application than SCSI.

We run the INGRES rdbms here, and still find that we're compute bound, that we get significant improvement by upgrading the cpu. We just put an SS2 in place of an SS1 for our test db work, and my benchmarks show basically the slightly better than twice as fast mips ratings for the machines. Oracle, I understand is even more cpu-bound (if there can be such a thing :-).

>o We have heard rumors that Sun will be using the 4/490 as it's sole platform > for newer and more advanced computing ie: multiprocessing. If this means > that the 4/490 will be upgradable, it would make good strategic sense for > us to purchase one.

Strategy, unfortunately, can't really be tested by anything but time :-)

>[..]

I realize that scsi is somewhat slower than the IPI, but I'd be interested in seeing what the numbers really are. We're faced with a somewhat similar situation here, and I'm leaning toward the multiple SS2s with a split of the apps as my solution. The other advantage of the scsi drives is that you can migrate them to another platform. The RS/6000 and the new fast HPs all use scsi drives.

--

From: Tom Ploegmakers <opal!uunet!nikhefk.nikhef.nl!tom> Organization: Nikhef-K, Amsterdam (the Netherlands).

hi,

your last point about the 4/490 is important. i can't say much because we signed a nondisclosure, but you should talk to some good sales reps about it.

We are looking for replacement of our gould pn9080 as well as additional horsepower. We are thinking about one big server for file and user service. We need to service some 50 users that do fysics work. Half of them will use a workstation. The other half is to use (grafics-) terminals. The main server has to give part of the cpu service as well. But a big part has to come out of the workstations. We intend to run NQS and keep them busy 24 hours a day.

We run suns in diffeent flavors already so we are looking at new suns under the sun. But silicon grafics and encore as well as cdc are also in the picture.

Hope you can use it, good luck.

--

From: Mike Nesel ames!elxsi.dfrf.nasa.gov!nesel@uunet

I agree, it seems like either way will have advantages for you. Yes, the 490 is the basis for the Sun multiprocessor due Real Soon Now.

Why not go with the multiple Sparcs, as this seems to be your inclination, anyhow? It is more flexible, and you can play with some state of the art distributed computing software toys.

My $0.02

--

From: opal!uunet!sirius.astro.uiuc.edu!hr (Harold Ravlin)

>We are faced with the decision on whether to buy a Sun 4/490 server or >get two or three Sparcstation 2's with multiple 1GB SCSI disks.

>o It seems to me that multiple Sparc 2's would buy us more aggregate compute > power than a single 4/490. While the 4/490 provides greater i/o capacity

A single SS2 is faster computationally than a 4/490.

>o Most of our network bandwidth is consumed by NFS traffic

The 4/490 has a higher I/O rate than the SS2, but does it matter? (I don't know). My guess is that the ethernet is the bottleneck followed by disk seek times.

>o Sparc 2's are cheaper and can always be relegated to a user's desk... Or upgraded to 'SS3s' or whatever comes later.

>Here are some points that favor the 4/490.

>o Sun sales is offering us a _substantial_ discount for trading in our old > 3/280 server...

Sun has stated that they will support the Sun 3s for 5 years, albeit at dropping (and more expensive) levels. One problem with multiple architectures is that you have to have two of every program.

>o Maintenance on the 4/490 is comparable to the maintenance we pay for > our 3/280.

But how does it compare to a SS2?

>o We have heard rumors that Sun will be using the 4/490 as it's sole platform > for newer and more advanced computing ie: multiprocessing. If this means > that the 4/490 will be upgradable, it would make good strategic sense for > us to purchase one.

Rumors that I've heard are that Sun should have a multiprocessor board for the 4/490 out by the end of the year rated at serveral bOPS (bunches of Operations/Sec).

The SS2 supports SBUS cards like the SPARCPrinter inteface. The 4/490 supports VME bus cards like FDDI interfaces. However, I've heard that there will be an FDDI interface for the SBUS and a rumor that Sun may have an SBUS on a future incarnation of the 4/490.

My suggestion would be to assume that either path (4/490 or SS2s) would do the job. Look at how much it costs to keep them going. Incrementally, hardware costs are cheaper on SS2s, but in sufficient numbers will exceed the price of a 4/490. (Mechanical devices like tape drives are the big drains in either case). Also, two SS2s will cost more for software support that on 4/490.

Still confused? So am I and we have both a 4/490 and a SS2.

--

From: eplrx7!mcneill@uunet (Keith McNeill) Du Pont Company

Part of the decision depends on what Sun's future plans are for both platforms. Pull some of your UPS muscle & make Sun give you a non-disclosure talk to give you some ideas on what is ahead. Don't make a decision in the dark! There are some things on Sun's horizon which might make you go one way or another.

--

From: opal!uunet!ee.ecn.purdue.edu!curt (Curt Freeland) Purdue University Engineering Computer Network

[stuff deleted]

>o We have heard rumors that Sun will be using the 4/490 as it's sole platform > for newer and more advanced computing ie: multiprocessing. If this means > that the 4/490 will be upgradable, it would make good strategic sense for > us to purchase one.

Sun is set to announce a new server this summer (roughly 3 months away from what I have heard). Rumour has it that this will be a multi-processor capable system with the TI Viking chip (60 Mips Sparc). The CPU board will allow you to install up to 8 of these chips. Sun is very good at having a "fire sale" before a new product is released... so I suspect this is the reason behind the substantial discount. From all I have heard, the 4/490 backplane/memory will be used in the new system. A swap of processor boards could convert a 490 to one of the new boxes.

It might be worth your while to ask Sun about the new server, and what the projected prices are. We are awaiting a non-disclosure on it. All I have heard has come from the trades, and a few inside sources at Sun. If a single CPU system was (price-wise) comparable to the 4/490, and it has all of the I/O caching, and 64 contexts, it might be worth your while to get one of those (60 MIPS vs 22 MIPS).

--

From: opal!uunet!cs.vu.nl!sater (Hans van Staveren) Vrije Universiteit Amsterdam, Holland

I cannot give you much input, since we are thinking about this same question ourself, but one point I think should be made. In general it is unwise to run applications on fileservers. This wrecks the cache.

We are tended to go for the SS2, especially since there is an attractive trade in program for our 3/50's connected to it.

By the way, if you are interested in FDDI, rumours have it that it will be on SBUS before the end of the year.

If you encounter any interesting arguments pro or con, I would be interested.

--

From: opal!uunet!ECN.NL!bernards Organization: Netherlands Energy Research Foundation ECN

Hah, we have an old 4/280.

My Nephew who works at Sun NL gave a lecture about NFS at the SUG-NL Member's meeting.

Using a few SS2's could be a low cost NFS serving alternative for the expensive 4/490 server. Basically due to Ethernet tranceiver limitations. I don't know if his boss liked that story ;-)

His story covered almost all your considerations you made

I'm writing a proposal for some system upgrades and I include some of his _and_ yours and some other guys into that.

The main reason why we're thinking about using SS2's instead of one big Sun is mainly NFS traffic on our backbone.

Keeping user stuff on a local SCSI is a big win.

Our 23 Discless 3/50 (do not laugh, on One 4/280 ) saturates the net heavily. SCSI disks are becoming very cheap nowadays (~DFL 1000,- per 100MB) SMD and IPI disks are not as cheap as SCSI's

On a SS2 server the suggestion is to add a SCSI Sbus controller per Disk just for boosting throughput. A SS2 can hold up to 4 SCSI controllers (one on board, 3 on Sbus)

Using the new Wren VII 1.2 Gb disk can get you about 4.4 Gb when using 1 disk per SCSI controller. You can quadruple this if you are using as a NFS readonly server (Software packages central installation)

Hope this is usefull info.

--

From: (Austrailian Poster wished to remain anonymous)

All I have is speculation on the future. I *suspect* the 4/490 is the next place Sun will push performance if only because their market looks a bit funny at the moment. Between that and the advantage of getting rid of your 3/280 I like that option. Mind you, we're more than happy with our new SS2s, they have 3 background cpu jobs running and still don't feel too ugly to use. On the SCSI/IPI option, are youlimited to Sun support? If not I remember reading it was possible to put 3rd party SCSI on the VME bus.

--

From: opal!uunet!indetech.com!tilman (Tilman Sporkert x2084) Organization: Independence Technologies, Inc. Fremont, CA

Wow, what a long article. [ Wait `till you see the summary! :-) -gb)

Some general comments on some of your issues (I used to be the Computer Systems Manager at Cypress Semiconductor, and therefore have quite some experience with these things):

Cost: No way that you can beat the Sparc2 solution with any 4/490 configuration. For all SUN equipment, you pay a lot for the enclosure! Also consider the space you need. A 4/490 has to go in a machine room, a Sparc2 can be somewhere on some desk in some corner.

Performance: Yes, the 4/490 has better I/O performance. But you can buy more than one SS2 for the same money.

Upgrade: Sun's upgrades are usually very, very expensive. For example, I once bought a 4/390 because I needed a big database server with lots of ASCII ports real quick, and the 4/490 wasn't ready yet. Well, a few month later, the 4/490 was there. At the same time, they cut the price for disk drives in half. Bottom line was the 4/490 in the same conf. as my 4/390 was only $10000 more expensive (machine price around $100,000). But the upgrade was $40000!!!

Disks: If you have a 4/490, you can stuff 4 1GB IPI drives into it. End of the story. More drives: buy an expansion cabinet for very big bucks! However, you can always stack SCSI disks on top - but then you could start with an SS2 to begin with!

Don't make the mistake of using your NFS server for CPU jobs as well - NFS would compete with that job for system resources, resulting in paging and reduced NFS performance for everybody.

--

From: opal!uunet!att!mtunf!rgh Organization: AT&T BL Middletown/Lincroft NJ USA

In article <1207@gaboon.UUCP> you write: > >We are faced with the decision on whether to buy a Sun 4/490 server or >get two or three Sparcstation 2's with multiple 1GB SCSI disks. > == Stuff deleted ==

Just remember that the Sparc 2 has about 2 times the compute power of the 490...If you get a 490, your users are almost certainly better running their simulations on something other than the server which will be pretty busy just doing the NFS stuff.

--

From: opal!uunet!ficc.ferranti.com!peter (Peter da Silva) Organization: Xenix Support, FICC

Not anything to do with computers...

> Gary Blumenstein United Parcel Service, Research & Development

You guys really need to work hard to fight off those Federal Express flying trucks.

--

From: opal!uunet!att!cbnewsh!rvk (R. Kline)

> We are faced with the decision on whether to buy a Sun 4/490 server or > get two or three Sparcstation 2's with multiple 1GB SCSI disks.

we have a 4/490 and a Sparcstation 2 (ss2) in our department. the 4/490 is a server, and the ss2 is used as a workstation, so I cannot give an apples and apples comparison. the biggest attraction for the 4/490 as a service is that it was designed to be a transaction processing machine. you can check an expert for details, but the essence is that the 4/490 has about 64 context switching registers versus a couple for the ss2 (again, check an expert). the 4/490 has 64 bit buses.

even though the SPECmarks for the two machines are essentially the same, the I/O for the 4/490 is designed to support lots of users. the 4/490 seems very peppy, although this is unfair because the ss2 is on a desk, and must access a server itself, putting it at a slight disadvantage.

for simulations where I/O is not an issue, I suspect the two machines would perform almost identically. as servers, it's well to keep in mind that serving functions have top priority, and if the machines are both running simulations and acting as servers, I/O is apt to be important.

in the end, since you're talking serveral ss2's versus one 4/490, it's even harder to say really, beyond the obvious that the raw CPU performance of the ss2 and the 4/490 are about the same. my guess is that server performance and CPU performance are quite different things, and while the ss2 will perform as a server, the 4/490 was designed to be one.

--

From: opal!uunet!terre.DMI.USherb.CA!beauchem (Denis Beauchemin) Organization: Sisca Informatique, Inc.

Hi Gary,

I can tell you right away that the current 4/490 will be replaced by a more powerful machine within 1-2 months. More precisely, the CPU card will be replaced by a 50 MHz version similar to the one currently in use by the SS2. There will also be multiprocessing capabilities built into the machine. So the need for more power should be served by it.

The 490 offers more expansibility than the SS2s, but its peripherals cost more money. It also requires (at least in Canada) a 220 Volt feed, which isn't always available.

All current models of 490 will be able to upgrade to the MP models at reasonable costs (according to Sun).

Hope this helps!

--

From: opal!uunet!cs.tut.fi!kv56962 (Kari Vaaranens) Tampere University of Technology, Tampere, FINLAND kjv@tampella.fi

I'd say, buy the 4/490! I have following reasons to my opinion:

-4/490 is more expandable -you can add more memory -faster disks (IPI) -upgrades to multiprocessor machines?

I'd bet my years salary that Sun will start selling both multiprocessor machines and upgrades for 4/490 as well as other servers. I'd bet too that there will be even faster disks than those 6 MBps disks.

I'd guess that those new models will be released at the same time as SunOS5 (SYSVR4) which is going to be out this summer! The SYSVR4 is the first SunOS which has multiprocessor support. So, it might be a good idea to wait a while... Then again why wait if you need the computation power now!

I don't understand, why accessories for 490 should be more expensive than for SS2. That is the tape drives, CD-drives etc. Besides Sun gives you a free CD-drive when you buy a server! (I suppose you mean workstations when you mention SS2? They are available as servers too, you know.)

When it comes to disks, try Seagates 2.5 GB IPI disk (or Fuji disks)... On the other hand, they are third party disks and Sun support may not like them...

-- Gary M. Blumenstein garyb@gallium.UUCP



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:14 CDT