From: | Brian Ghidinelli <brian(at)pukkasoft(dot)com> |
---|---|
To: | sfpug(at)postgresql(dot)org |
Subject: | Followup on benchmarks, results |
Date: | 2006-01-17 23:17:44 |
Message-ID: | 43CD7B18.8030108@pukkasoft.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | sfpug |
I just tested the 4-disk RAID-10 for data and 2-disk RAID-1 for
OS/transaction log. I'm really surprised by the speed increase but
perhaps it caters to the type of test pgbench performs? These are
transactions per second excluding connection overhead:
c=1 c=2 c=5 c=10
304.2 393.8 360.4 319.1
That represents between a 37-60% improvement over using 2 separate
2-disk RAID-1 arrays (one for data, one for log). I know this isn't all
that interesting to those who play with big scale hardware and drive
arrays but I think 4 or 6 disks in a box a pretty common setup for us
little guys. Here are my full results:
<http://www.ghidinelli.com/workspace/postgres/postgres_20060116.html>
After just a little testing, it becomes quickly obvious that you could
spend years doing this. :)
Cheers,
Brian
From | Date | Subject | |
---|---|---|---|
Next Message | Quinn Weaver | 2006-01-21 22:28:56 | CodeCon coming up; interesting PostgreSQL talk |
Previous Message | Brian Ghidinelli | 2006-01-17 19:12:31 | Re: Benchmarking results, questions |