>>>>> Badari Pulavarty (BP) writes:
 BP> Sure. I think it will improve the allocation case. 
 BP> Non-allocation case, should be pretty much same, provided
 BP> I got contiguous layout on the disk. Isn't it ?


not allocation only:

[EMAIL PROTECTED] root]# time /work/tests/fwrite /test/fff 64 10000
real    0m13.102s user    0m0.027s sys     0m4.003s

[EMAIL PROTECTED] root]# time /work/tests/fwrite /test/fff 64 10000
real    0m22.210s user    0m0.006s sys     0m3.241s

[EMAIL PROTECTED] root]# time /work/tests/fwrite /test/fff 64 10000
real    0m23.242s user    0m0.005s sys     0m3.202s

[EMAIL PROTECTED] root]# time /work/tests/fwrite /test/fff 64 10000
real    0m24.777s user    0m0.011s sys     0m3.175s

I think at least couple things're involved:
 - for already allocated block I don't open a transaction
 - extents code caches last found extent, so most lookups
   just do few compares with in-inode integers and don't
   walk though a tree

btw, I've observed the following issue: sometimes vm can
run several ->writepages() against a same file or mix
->writepages() with ->writepage(). I think this can break
layout ...

thanks, Alex

-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to