Re: File writing performance
From: John Dallman (jgd_at_cix.co.uk)
Date: 10/28/03
- Next message: Paul Landay: "Re: ipfx package for AIX 5L?"
- Previous message: Steve Nottingham: "Re: AIX 5L 5.2 in RS/6000 F40"
- In reply to: Lu: "File writing performance"
- Next in thread: Shea Martin: "Re: File writing performance"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Tue, 28 Oct 2003 22:57 +0000 (GMT Standard Time)
In article <89ee3c55.0310281428.5f4a3eaf@posting.google.com>, qlu@lgc.com
(Lu) wrote:
> The weird thing is, it takes much longer elapse time (wall-clock time)
> than CPU time (e.g., 20 min vs. 2 min) in IBM-AIX (AIX4.3), while in
> all other systems, these two kinds of time are almost the same (e.g.
> 2min). The extra waiting in IBM machine does happen during file
> pointer locating and data writing stage. In all cases, this program is
> the only runing application during testing.
Look carefully at the file buffering defaults in the various systems' C
run-time libraries, and in the operating system configurations.
I had something like this that bit a Windows program, where unbuffered
writing to network disk ran at least ten times slower than local disk on
one version of Windows, but not on an older one. The newer version was
waiting for the round trip to finish and didn't return to the calling
program until the data had really hit disk on the server. The older one
claimed OK once the OS had hold of the data, and accumulated multiple
write calls until it had a disk cluster's worth. Something a bit like that
could be happening to you.
---
John Dallman jgd@cix.co.uk
- Next message: Paul Landay: "Re: ipfx package for AIX 5L?"
- Previous message: Steve Nottingham: "Re: AIX 5L 5.2 in RS/6000 F40"
- In reply to: Lu: "File writing performance"
- Next in thread: Shea Martin: "Re: File writing performance"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|