Hi Yu > 1.Why the performances under different subdirs varies so much? for example > in /mnt/test/0 the performance is 72/80, while in /mnt/test/1 the > performance is 49/53 I think I agree with Thomas that physical data location does matter here. Also, do you reformat the partition before every iteration of the test? > 2.The extents of all files are nearly the same, but their performances are > different. Are there other factors that influence the performance beside of > the extents(fragmentation) of the file? Ehm, I/O allocator? block allocator (i think orlov allocator is the default, but I don't if it's not the case anymore) > 3.Is it true that more files exist in a directory, lower performance? for > example the performance in /mnt/test/5/tmpfile is 66/75, while the > performances in /mnt/test/5/tmpfile2 and tmpfile3 are 57/64 and 54/59 maybe it's a matter of dentry lookup, but once it's resolved, it will be cached in dcache. So maybe it's a matter of icaching, since 5G file (am I right) could take a lot of icaching (depends on block size too). Final thought, have you considered to use non journalling mode? I mean, i remember ext3 has 3 mode of journalling, pick the weakest one and try to reduce flush out time. regards, Mulyadi -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
