On Thursday, October 25, 2001 03:06:27 PM -0600 Eric Whiting <[EMAIL PROTECTED]> wrote: > Here is some feedback of 2.4.13+the patch from Chris. Two tests: local > fs and NFS. I still see odd things happening at files above 300G. This > is reiserfs formatted -v2 (3.6) with a default mount (tail). > > Linux box is 512M RAM. The files fit in buffer cache. I run the bonnies > of size n+50. Results are fairly constant until I hit about 300M. Then > they seem to fall off an edge. I'll include only two sample points for > each setup. Bonnie really needs 2 * RAM to be reliable. Notice how on the 100M test you've got 99% CPU usage? That's most likely because the 100M test is doing nothing but hammering on the various caches. The drop off point comes at 300M because the test starts taking long enough for kupdate and friends to jump in an throttle the writes. How long does each phase take (esp the intelligenty write phase)? -chris
