| Lists: | pgsql-bugs |
|---|
| From: | pgsql-bugs(at)postgresql(dot)org |
|---|---|
| To: | pgsql-bugs(at)postgresql(dot)org |
| Subject: | PostgreSQL BugTool Submission |
| Date: | 2000-08-24 04:49:03 |
| Message-ID: | 200008240449.e7O4n3603230@hub.org |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-bugs |
Doug Mitchell (pgsql(at)dougmitchell(dot)com) reports a bug with a severity of 3
The lower the number the more severe it is.
Short Description
Excessive seeking by backend
Long Description
The backend call lseek much more often than it has to. The storage manage should not do either of the following:
1. Repetetive seeks on the same location without any intervening read or write.
2. Seeking to the next block to read when it is already the current
location.
These excessive system calls hurt performance by forcing a full context switch to kernel mode everytime.
Sample Code
I did a combination of sequential scans, index scans, and joins while watching the backend with strace under Linux 2.2. The results are really rather surprising.
No file was uploaded with this report
| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Doug Mitchell <pgsql(at)dougmitchell(dot)com> |
| Cc: | pgsql-bugs(at)postgresql(dot)org |
| Subject: | Re: PostgreSQL BugTool Submission |
| Date: | 2000-08-24 05:16:23 |
| Message-ID: | 25636.967094183@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-bugs |
pgsql-bugs(at)postgresql(dot)org writes:
> The backend call lseek much more often than it has to.
Denis Perchine fixed this a couple of months ago. You might care to
repeat your experiments with current sources (see our CVS server, or
the nightly snapshot tarball on our FTP server). If there's still
a problem we'd like to know before it's too late to tweak 7.1 ...
regards, tom lane