Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 9 Aug 1995 14:10:47 +0200
From:      esser@zpr.uni-koeln.de (Stefan Esser)
To:        David Greenman <davidg@freefall.freebsd.org>
Cc:        CVS-commiters@freefall.freebsd.org, cvs-sys@freefall.freebsd.org
Subject:   Re: cvs commit: src/sys/ufs/ffs ffs_inode.c
Message-ID:  <199508091210.AA09425@Sysiphos>
In-Reply-To: David Greenman <davidg@freefall.cdrom.com> "cvs commit: src/sys/ufs/ffs ffs_inode.c" (Aug  3, 22:49)

next in thread | previous in thread | raw e-mail | index | archive | help
On Aug 3, 22:49, David Greenman wrote:
} Subject: cvs commit: src/sys/ufs/ffs ffs_inode.c
} davidg      95/08/03 22:49:18
} 
}   Modified:    sys/ufs/ffs  ffs_inode.c
}   Log:
}   Use the correct flags (IO_SYNC -> B_SYNC) when deciding to do a sync or
}   async write in the section that changes the filesize. The bug resulted
}   in the updates always being async.

What performance improvement is 
to expect from that patch ?

Does it only make a difference if
a file is modified, not if it is
created or deleted ?

I did some tests, creating and 
deleting 1000 files from a small
shell script:

#!/bin/sh
for i in 0 1 2 3 4 5 6 7 8 9
do
    mkdir $i
    for j in 0 1 2 3 4 5 6 7 8 9
    do
        mkdir $i/$j
        for k in 0 1 2 3 4 5 6 7 8 9
        do
            echo "file $i$j$k" > $i/$j/$k
        done
    done
done


I found that the "-async" option 
made no difference at all !

Using "simple" tags instead of 
"ordered" tags for writes made the 
drive reorder file and meta data 
updates, resulting in 30% higher
file creation rate ...

f1000.sh
========

tags    simple          ASYNC           SYNC
---------------------------------------------
4       n               37.59           36.65
16      n               36.89           31.75
16      y               28.55           28.58

rm -rf *
========
tags    simple          ASYNC           SYNC
---------------------------------------------
4       n               24.30           23.34
16      n               23.87           21.53
16      y               22.56           22.82

I would have expected the ASYNC 
mount to make much more of a 
diference ...

Regards, STefan

-- 
 Stefan Esser				Internet:	<se@ZPR.Uni-Koeln.DE>
 Zentrum fuer Paralleles Rechnen	Tel:		+49 221 4706021
 Universitaet zu Koeln			FAX:		+49 221 4705160
 Weyertal 80
 50931 Koeln



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199508091210.AA09425>