| From: | Gregory Stark <stark(at)enterprisedb(dot)com> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | Greg Smith <gsmith(at)gregsmith(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com>, ITAGAKI Takahiro <itagaki(dot)takahiro(at)oss(dot)ntt(dot)co(dot)jp>, Postgres <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: posix_fadvise v22 |
| Date: | 2009-01-02 20:40:49 |
| Message-ID: | 878wptv3u6.fsf@oxford.xeocode.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
> In principle you should be able to adjust the constant so that vmstat
> shows about 50% CPU busy, and then enabling fadvise should improve
> matters significantly.
I think in practice individual queries don't interleave much cpu with i/o
work. A single random page fetch is 5ms which is an awful lot of cpu cycles to
be sinking somewhere. In practice I think this is going to be single-digit
percentages.
Aside from big sorts and such which tend not to be interleaved with i/o the
main time I've seen queries have a significant cpu load is when the data is
mostly in cache. In which case prefetching would be hard pressed to help at
all.
We could construct a synthetic case but the main point of this feature is to
make use of raid arrays that are currently going idle, not to pick up a few
percentage points for single spindle systems.
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
Ask me about EnterpriseDB's On-Demand Production Tuning
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Bruce Momjian | 2009-01-02 20:43:41 | Re: posix_fadvise v22 |
| Previous Message | Robert Haas | 2009-01-02 20:35:18 | Re: Significantly larger toast tables on 8.4? |