xattr performance and Samba4

[tridge at samba.org]

Bjoern,

 > there has been another ACL Test some time ago been made by Andreas 
 > Grünbacher, this one says that JFS is much better than JFS looks in 
 > your tests.

Those tests were with a 2.4.20 kernel, which is quite different from
2.6.10 that I tested with.

 > One of the reasons which might explain it's good ACL/EA 
 > performance (which can't be seen in your benchmarks ;) )

I think you may have missed the graph I did of dbench performance with
xattrs enabled (xattrs are the same as EAs).

  http://samba.org/~tridge/xattr_results/xattr.png

It might be that jfs somehow magically multiplies its performance by
10x when a posix ACL is added to a file, but I doubt it.

I suspect the really low jfs results in my tests are just a bug, and
hopefully will be fixed. Otherwise jfs is doomed on linux.

Cheers, Tridge