Date: Wed, 26 Apr 2000 23:47:44 -0400 From: Mike Tancsa <mike@sentex.net> To: Matt Dillon <dillon@FreeBSD.ORG> Cc: stable@FreeBSD.ORG Subject: Re: cvs commit: src/sys/gnu/ext2fs ..... Message-ID: <4.2.2.20000426234307.0342e878@mail.sentex.net> In-Reply-To: <200004262036.NAA81630@freefall.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
At 01:36 PM 4/26/2000 -0700, Matt Dillon wrote:
>dillon 2000/04/26 13:36:37 PDT
> MFC sequential write detection heuristic. This extends the sequential
> read heuristic to also cover sequential writes, causing write-behind
> to be used only for the sequential write case. This solves a number of
> performance issues with random writes to medium sized files using
> large block sizes, as occurs with DBM files.
Well, it does seem to make a difference for me. I was building another
box, and decided to try out a 'before' and after.
The first is from a stable box from the AM, the second with the above commit.
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
500 17721 84.5 18474 34.8 6166 12.4 14853 97.6 25036 30.6 110.7 2.0
500 17566 82.9 18419 35.0 8301 16.9 14885 97.1 25014 20.5 108.2 1.6
Sure enough, an increase in the rewrite performance on the drive.... The
other differences are most likely statistically insignificant, but the
rewrite does seem to be valid. Way to go!
drive is a
ad1: 39082MB <Maxtor 94098U8> [79406/16/63] at ata0-slave using UDMA33
on a
Intel PIIX4 ATA33 controller
---Mike
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4.2.2.20000426234307.0342e878>