Filesystem and TCP perform poorly?

[Mike Cheponis]

http://bsd.slashdot.org/article.pl?sid=03/11/05/1536226&mode=thread&tid=122&tid=185&tid=190&threshold=-1:


`Research conducted at MIT [mit.edu] found BSD's filesystem implementation
to be "very poorly performing." [1] Even BSD's acclaimed TCP/IP stack
has lagged behind, according to this study. [2]'

[1] http://web.mit.edu/tytso/www/linux/ext2intro.html

[2] http://cs-tr.cs.rice.edu/Dienst/Repository/2.0/Body/ncstrl.rice_cs/TR99-335/postscript

[Charles M. Hannum]

Post by Mike Cheponis
http://bsd.slashdot.org/article.pl?sid=03/11/05/1536226&mode=thread&tid=122
`Research conducted at MIT [mit.edu] found BSD's filesystem
implementation to be "very poorly performing." [1] Even BSD's acclaimed
TCP/IP stack has lagged behind, according to this study. [2]'
[1] http://web.mit.edu/tytso/www/linux/ext2intro.html
[2]
http://cs-tr.cs.rice.edu/Dienst/Repository/2.0/Body/ncstrl.rice_cs/TR99-335
/postscript

Both of these papers are very old and have no bearing on reality any more.
Bringing them up now is nothing short of FUD.

[Manuel Bouyer]

Post by Mike Cheponis
`Research conducted at MIT [mit.edu] found BSD's filesystem implementation
to be "very poorly performing." [1] Even BSD's acclaimed TCP/IP stack
has lagged behind, according to this study. [2]'
[1] http://web.mit.edu/tytso/www/linux/ext2intro.html

This is a very old article.
Basically ext2fs is FFS without fragments (just look at sys/ufs/ext2fs in
NetBSD sources). If I remember properly, the article was comparing
both filesystems in default config, which means ext2fs was async and
ffs semi-sync.
Also I think the system used to benchmark FFS didn't have R/W clustering
at this time.
And, last, at this time linux has his dynamic sized buffer cache, while
the system used to benchmark FFS still had the static fixed-size buffer
cache.
The difference in performances were related to both default settings,
and others differences in the operating system. The BSD has done lots of
progress since then :)