Tested the flusher slightly

OGAWA Hirofumi hirofumi at mail.parknet.co.jp
Sun Jun 23 08:54:21 PDT 2013


I've tested the flusher slightly to see some tweak has any difference.

Pentium D (2 cores), 3GB memory, SATA HDD (no TCQ, UDMA133)
[Yeah, this is native machine, not kvm]

Test is "dd if=/dev/zero of=/mnt/file bs=4K count=1M" with following

1) Use WRITE command for save_sb()
2) Use WRITE_SYNC command for save_sb()
3) Add FLUSH command before save_sb()

As expected, for now, tux3 is slightly faster than ext4. Maybe, because
we are using the write-optimized block allocation(?).

And FLUSH command makes slower slightly. blocksize 512 makes slower, and
increase cpu systime, because probably buffer_head overhead, especially
sort of buffer_head (perf says buffer_index_cmp() is top cpu user).

4294967296 bytes (4.3 GB) copied, 64.511 s, 66.6 MB/s

real	1m4.607s
user	0m0.428s
sys	0m9.580s

[4096 blocksize,WRITE]
4294967296 bytes (4.3 GB) copied, 63.943 s, 67.2 MB/s

real	1m3.960s
user	0m0.276s
sys	0m7.832s

[4096 blocksize,WRITE_SYNC]
4294967296 bytes (4.3 GB) copied, 63.821 s, 67.3 MB/s

real	1m3.834s
user	0m0.204s
sys	0m7.916s

[4096 blocksize,WRITE_SYNC,FLUSH]
4294967296 bytes (4.3 GB) copied, 64.006 s, 67.1 MB/s

real	1m4.016s
user	0m0.196s
sys	0m7.840s

[512 blocksize,WRITE_SYNC]

4294967296 bytes (4.3 GB) copied, 71.6808 s, 59.9 MB/s

real	1m11.755s
user	0m0.384s
sys	0m20.048s

OGAWA Hirofumi <hirofumi at mail.parknet.co.jp>

More information about the Tux3 mailing list