Re: Oracle on v40z vs Oracle on v490

Hi!

    Unfortunately, we don't have any possibility to get a test box with Sparc IV+ and try any performance tests in such a short time (three weeks). So we have to make a decision on the basis of our "feelings" and other's experience. But according to replays here (your and Billy), AMD boxes look more preferable...

We've tested v20z already - in our tests the "core multiplicator" was between 1.5 and 2.

Alexey Sergeyev

"NetComrade" <netcomradeNSPAM_at_bookexchange.net> wrote in message news:4360fc9b.1467958468_at_localhost...
> On Thu, 27 Oct 2005 14:09:01 +0400, "Alexey Sergeyev"
> <saefido7_at_devexperts.com> wrote:
>
>> At the moment we have several RAC systems (Sun Fire V440, 4xUltraSparc
>>IIIi-1062 MHz each, Oracle 10.1.0.4, up to 6 nodes). We are going to
>>upgrade
>>that hardware, and have now two options: to use Sun Fire V40z with
>>dual-core
>>AMD Opteron CPU, or to replace theses boxes with Sun Fire V490 with
>>UltraSparc IV+.
>>
>> Some time ago we tested Oracle on V20z (2xAMD 875, 2.2 GHz) in
>>comparision with our V440s, and found out, that AMD system is almost three
>>tims faster than a v440. So, if we use v40z with 4xAMD 880 (2.4 GHz), we
>>can
>>expect up to 5.4 times performance improvement (3 times per core, and
>>multiply by 1.8 - the second cores "weight").
>>
>> But we know nothing about Oracle performance on UltraSparc IV+. SUN
>>says, UltraSparc IV+ 1500 is UP TO 5 times faster than UltraSparc III.
>>I've
>>traied to find any real examples, but didn't find any... So i have two
>>questions (really two blocks of questions :)) )
>
> My experience with these machines has nothing to do with RAC.. but..
>
> Back when I was looking to replace our aging Sparc II E4500, based on
> per CPU calculations from www.spec.org, it would seem that v40z
> (single core 852) would almost 8x faster (per CPU), other machines
> would lag behind.
>
> Sun Microsystems Sun Enterprise 3500/4500 198 212 100%
> Sun Microsystems Sun Fire V40z (Opteron 852) 1558 1741 787%
> Sun Microsystems Sun Fire V880 (1200MHz) 625 700 316%
> Dell PowerEdge 6600 (3.0 GHz Xeon MP) 1379 1408 696%
> Dell PowerEdge 3250 (1.5GHz/6MB, Itanium2) 1099 1099 555%
>
> So we replaced 10 CPU Sparc II machines, by 4CPU AMD machines.. Even
> if v40z's were only 6x faster, v40z was supposed to behave like a
> 24CPU sparc II (10cpus 6x faster * 4/10 CPUs)
>
> I must say we are extremely happy with 3xv40z's that we installed,
> together with an EMC CX500 array (although EMC is still evil). Server
> loads never go beyond 1-2, unless some batch jobs are running, batch
> job time completion has increased by a factor of X (5-10 times), and
> we didn't have a 'load' problem once due to CPU/memory bottlenecks yet
> (and Sept is one of the busiest times of the year for us, since we are
> a sports-related website, db is mostly "OLTP")
>
>>
>> If reliability and performance of an AMD system will be, at least, not
>>worse, than the mentioned Sparc system, then v40z seems much more
>>preferable
>>for us - $40k per box looks much better than $90k for one UltraSparc box.
>>
>
> Also, I am not sure if your 'dual core' calculations are correct. For
> Sparc IV Sun it was bet 1.7 and 1.8, for AMD's it might be smaller, I
> remember reading 40-50% performance gain.
>
> Anyway, I haven't seen any CPU ratings on Sparc iV+.. I kind of doubt
> they're 5x faster than Sparc III, but do also take into consideration
> Oracle pricing together with your hardware calculations (you want the
> fastest CPU's, that's why we went with single core)
>
> I strongly advise getting a Sparc IV+ test box, or to get spec metrics
> from Sun.
>
>
> .......
> We use Oracle 8.1.7.4 and 9.2.0.6/7 on Solaris 2.7 and RH4
> remove NSPAM to email
Received on Fri Oct 28 2005 - 05:32:05 CDT