Chris, It 'feels' faster on my NFS bonnie benchmarks. I still see some degradataion in speed on larger files -- but I think that is a different 'feature' of the filesystem.
512M RAM 2.4.11pre1 NFSV3 server lightly loaded ReiserFS version 3.6.25 (v2 format filesys) 100M network -- solaris 8 client 100M file Writing with putc()... done: 10294 kB/s 59.3 %CPU Rewriting... done: 7312 kB/s 12.4 %CPU Writing intelligently...done: 11099 kB/s 14.4 %CPU Reading with getc()... done: 7898 kB/s 37.5 %CPU Reading intelligently...done: 132682 kB/s 99.8 %CPU 400M file Writing with putc()... done: 6528 kB/s 44.7 %CPU Rewriting... done: 2465 kB/s 5.2 %CPU Writing intelligently...done: 6280 kB/s 15.5 %CPU Reading with getc()... done: 3643 kB/s 18.2 %CPU Reading intelligently...done: 85817 kB/s 95.3 %CPU These speeds are still pretty good. I see a similar scale down in speed running the tests local on the 2.4.11pre1 box. notail doesn't seem to change the results much. eric Chris Mason wrote: > > Hello everyone, > > 2.4.11pre1 has some significant changes which affect almost everything that > touches a disk in linux. The buffer cache is now backed by an address > space per device (made possible by Andrea's blkdev changes in 2.4.10), and > many cleanups were done by Linus and Al Viro. > > In every test I've run so far, this is a speed gain for reiserfs, I'd love > to see further results. > > This is one of the more experimental prereleases in 2.4.x, don't run it on > systems with important mail messages, source code, work files, homework, > love letters, tax records or anything else not found on your favorite > distribution's media. I don't expect problems, but.... > > -chris