I would like to hear from DBA who tested the cluster version
in production site.
what's the scalaiblity of this new version?
if you have 4 cpu machine and you add another 4 cpu machine
- what's the power you get , 5,6,7 cpu ?
From: Golan Nahum
Subject: sclability of the cluster version
X-Mailer: WebNews to Mail Gateway v1.1t
Date: 26 Sep 2008 12:07:58 -0700
X-Trace: forums-1-dub 1222456078 10.22.241.41 (26 Sep 2008 12:07:58 -0700)
X-Original-Trace: 26 Sep 2008 12:07:58 -0700, 10.22.241.41
Xref: forums-1-dub sybase.public.ase.cluster:96
Article PK: 48378
Subject: Re: sclability of the cluster version
User-Agent: Pan/0.132 (Waxed in Black)
Content-Type: text/plain; charset=UTF-8
Date: 30 Sep 2008 09:55:03 -0700
X-Trace: forums-1-dub 1222793703 10.22.241.152 (30 Sep 2008 09:55:03 -0700)
X-Original-Trace: 30 Sep 2008 09:55:03 -0700, vip152.sybase.com
Xref: forums-1-dub sybase.public.ase.cluster:97
Article PK: 48379
I'm not a DBA (anymore, but I was once!), but I've been working with the
product for nearly a year now, and done several POC's so I'm going to
reply even though I work for Sybase :).
It's generally not realistic to expect linear scalability with any
cluster solution, whether it's Sybase ASE or "someone else" (if you don't
believe me, check tpcc benchmarks, they prove this out). So, you won't
get the same scalability with 4 machines @ 4 CPU's each as you will with
a 16-way SMP machine.
HOWEVER you may see increased scalability, depending on how the
application is used (and how it's accessed). Some customers that I've
worked with have seen good linear scaling, some haven't seen what they
had hoped to but in every case it depended on the application, and how it
was accessed / used.
Generally speaking the less cache-to-cache transfer, contention, and
conflicting updates you have, the better the scale-out. The more of that
you have, the less scale-out you'll get ... no matter who the DBMS vendor.
Others may have different views but that's what I've seen so far.