| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | pgsql-performance(at)lusis(dot)org |
| Cc: | "PGSQL Performance" <pgsql-performance(at)postgresql(dot)org> |
| Subject: | Re: Performance of pg_dump on PGSQL 8.0 |
| Date: | 2006-06-14 15:51:47 |
| Message-ID: | 26780.1150300307@sss.pgh.pa.us |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
"John E. Vincent" <pgsql-performance(at)lusis(dot)org> writes:
> I've watched the backup process and I/O is not a problem. Memory isn't a
> problem either. It seems that we're CPU bound but NOT in I/O wait.
Is it the pg_dump process, or the connected backend, that's chewing the
bulk of the CPU time? (This should be pretty obvious in "top".)
If it's the pg_dump process, the bulk of the CPU time is likely going
into compression --- you might consider backing off the compression
level, perhaps --compress=1 or even 0 if size of the dump file isn't
a big concern.
Another possibility if your LAN is reasonably fast is to run pg_dump on
a different machine, so that you can put two CPUs to work.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Scott Marlowe | 2006-06-14 16:44:10 | Re: Performance of pg_dump on PGSQL 8.0 |
| Previous Message | Jim C. Nasby | 2006-06-14 15:18:04 | Re: Precomputed constants? |