My original posting was this:
> I have some information on a RAM-based disk for Sun systems that
> claims to offer some dramatic performance improvements in performance
> for applications such as Oracle. Has anyone had personal experience
> with this board, good or bad?
Most of my replies were people wanting a summary of my replies. I've
included the more useful ones below. Thanks to everyone who replied.
We've decided not to get the board, mainly because of the price. It
is cheaper to buy more regular memory. No one really gave it an
overwhelming review either. Here are the replies:
> I don't think ram disks have a signifacent advantage for unix (but more
> physical memory and more swap space does.)
>
> You have to evaluate if the figures are lying or the liars are figuring...
>
> marty
> leisner@eso.mc.xerox.com leisner.henr801c@xerox.com
> I can saw that we just installed it on a Sparcstation 2 with 128MB
> of memory and the user (using Valid's Allegro application) saw
> a dramatic improvement!
>
> He said it was like night and day...
>
> However, we installed it not very long ago and haven't had a change
> to take our own stats.......
>
> So, I'm glad the users happy, but I want real numbers before I
> come out and say "buy one!".
>
> Peter Sivo
> Amdahl/Advanced Systems
> peter@key.amdahl.com
> I looked at using the Turboswap board but found that it was more expensive
> than buying main memory. If you have not maxed out main memory then I would
> look at buying more of that first.
>
> ---
> Nicholas C. Franco ncf@cray.com
> Cray Research, Inc.
> No, but I'd be interested in how they lobotomize pagedaemon to prevent
> it from swapping out their disk. :) Or do they use a hardware
> solution with dedicated ram?
>
> --
> Adam Shostack adam@bwnmr4.harvard.edu
> Systems Manager 617-732-7692
> Surgical Planning Lab, Dept of Radiology Fax 732-7963
> Brigham and Womens Hospital, Boston
> I first spoke with CERAM's sales person last June. In essence,
> the product looked promising enough for us to evaluate it.
>
> I asked for a 480 MByte TurboSwap box because I believed that
> its best use was on CPU servers which run large CAD jobs.
>
> We couldn't get our act together over the summer. But, from
> about September through December, it became evident that CERAM
> may not actually "have the product" at least in such a large
> size.
>
> A CERAM engineer told me that they were having some troubles
> getting product ready for our evaluation.
>
> Well, it's January now. No evaluation unit. I'm not sure
> what that means.
>
> Bill Shorter
> bill@aloft.att.com
> We have tested the CERAM board for about 1 1/2 months. Our application
> just wasn't using swap space or /tmp much so we didn't notice any significant
> performance differences. We are a map maker using ESRI's Arc/Info software.
>
> We had no problems installing the 160MB version on a Sparc 2. We also had
> NO problems with this machine during the time we were testing the unit.
> We were running SUNOS 4.1.2 on the machine.
>
> If your app swaps alot or you can use the tmpfs you could probably try
> it out without affecting your real work.
>
> If you want more info just drop me a line or give me a call..
>
> Patrick O'Shaughnessy UUCP: sun!sunkist!ohio!patrick
> Thomas Bros. Maps patrick@thomas.com
> 17731 Cowan PHONE: (714) 863-1984
> Irvine, Ca 92714 FAX: (714) 757-1564
> Yes, we had a board in house for about a month for testing. We ended
> up not buying it, though it did offer some performance improvement.
>
> Of course, whether or not you'll see a perfomance gain depends on the app...
>
> Here, we use GIS software. Creating large plots can be quite compute
> intensive, and with the CERAM board we noticed about a 20% increase in
> speed. In addition, the workstation was still usable throughout the
> entire process. Without the board, a SPARC 2 with 32 MB RAM and 100+
> MB swap becomes almost unusable during certain parts of the process. I
> can't give you a lot of specific details because I didn't do any of the
> testing. I just installed it, rebuilt the kernel, etc. This is just
> info I picked up from those using it.
>
> An application that creates a lot of temporary files could really
> benefit from this - creating files in RAM is a lot faster than on a
> disk.
>
> Installation and de-installation were easy. Scripts are supplied to
> accomplish the tasks. The hardware install is just as easy as
> installing any other S-bus device.
>
> It did do some funny things to the kernel: Some processes no longer
> worked once the board was installed. The fix was boosting MAXUSERS to
> 256. (!) The software worked again, but doing a pstat with MAXUSERS
> set that high gives some interesting results. The value of certain
> statistics were returned: "astronomical".
>
> BTW: We had two 80MB cards "piggybacked" for a 160MB RAM disk.
>
> For the cost, we just didn't think they were worth it. For the money
> they want, I can add normal RAM to half a dozen or more workstations.
>
> ----
> Jim Sidaris Voice: (714) 863-1984 ext. 593
> Thomas Bros. Maps FAX: (714) 757-1564
> 17731 Cowan
> Irvine, Ca 92714 E-mail: jsidaris@thomas.com
-- Peter Steele Unix Services Manager peter.steele@acadiau.ca Acadia Univ., Wolfville, NS, Canada B0P 1X0 902-542-2201 Fax: 902-542-4364
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:23 CDT