linux-kernel - Squashfs 4.0 performance benchmark

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <6f9b6d3e0906080705g3692f1b7l9c75322633ea6d29@mail.gmail.com>
Date:	Mon, 8 Jun 2009 22:05:00 +0800
From:	Wenqiang Song <wsong.cn@...il.com>
To:	linux-kernel@...r.kernel.org
Subject: Squashfs 4.0 performance benchmark

Hi folks,

I just did a performance test on Squashfs 4.0, similar to this one:
http://www.celinuxforum.org/CelfPubWiki/SquashFsComparisons

The result is quit good. But I don't understand why performance is
better on a loop back file than raw partition.

My test machine:
AMD Sempron(tm) Processor LE-1200, 2G memory

Ext4 uncompressed data: 2887704K
Squashfs 4.0 Image: 1256892K, generate by "mksquashfs . test.squash"

2.1.1 Directory Lookup performance
Time taken to perform "ls -lR --color=alawys | cat > /dev/null" on
filesystem mounted

First time, Ext4
real    0m13.896s
user    0m3.556s
sys     0m4.436s

Second time, Ext4
real    0m6.700s
user    0m2.676s
sys     0m2.816s


First time, Squashfs on raw partition
real    0m9.850s
user    0m2.268s
sys     0m6.712s

Second time, Squashfs on raw partition
real    0m4.487s
user    0m2.096s
sys     0m2.392s


First time, Squashfs loop back file on an Ext4 partition.
real    0m8.356s
user    0m2.104s
sys     0m6.112s

Second time, Squashfs loop back file on an Ext4 partition.
real    0m5.112s
user    0m2.116s
sys     0m2.992s

---------------------------------------------------------------------------------------------
2.1.2 Sequential I/O performance

Time taken to perform "tar cf - | cat > /dev/null" on filesystem mounted

Ext4
real    5m28.694s
user    0m3.816s
sys     0m29.966s


Squashfs on raw partition
real    1m23.226s
user    0m2.416s
sys     1m2.680s


Squashfs loop back file on an Ext4 partition

real    0m53.867s
user    0m1.820s
sys     0m42.611s


--------------------------------------------------------------------------------------------
2.1.3 Random I/O performance
Random access pattern generated by find . -type f -printf "%s %p\n" |
sort -g | awk '{ print $2 }' > /tmp/sort

First time, Ext4
real    33m21.603s
user    0m5.724s
sys     0m33.734s

Squashfs on raw partition
real    7m35.962s
user    0m4.192s
sys     5m40.973s

Squashfs loop back file on an Ext4 partition
real    5m33.566s
user    0m3.316s
sys     4m47.942s

----------------------------------------------same test with 256M memory
Squashfs on raw partition
real    10m25.518s
user    0m4.432s
sys     6m41.133s

Squashfs loop back file on an Ext4 partition
real    11m0.271s
user    0m5.156s
sys     6m59.942s


Wenqiang Song

--
有志者，事竟成，破釜沉舟，百二秦关终属楚
苦心人，天不负，卧薪尝胆，三千越甲可吞吴