Tested the flusher slightly

[OGAWA Hirofumi]

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

> Daniel Phillips <daniel.raymond.phillips at gmail.com> writes:
>
>> 99% of blockdev of blockdev speed sounds like something to be proud
>> of. What is media speed, as measured by hdparm -t?
>
> root at duaron (~)# hdparm -tT /dev/sdb2
>
> /dev/sdb2:
>  Timing cached reads:   1610 MB in  2.00 seconds = 805.10 MB/sec
>  Timing buffered disk reads: 172 MB in  3.00 seconds =  57.28 MB/sec
>
> Note, that test is write, and no fsync(2).

Re-measured with conv=fsync.

# sleep 3600 < /dev/sdb2 &
# blockdev --setbsz 4096 /dev/sdb2
# time dd if=/dev/zero of=/dev/sdb2 bs=4K count=1M conv=fsync
[blockdev,fsync]
4294967296 bytes (4.3 GB) copied, 72.6971 s, 59.1 MB/s

real	1m12.717s
user	0m0.240s
sys	0m5.728s

4294967296 bytes (4.3 GB) copied, 72.6532 s, 59.1 MB/s

real	1m12.904s
user	0m0.368s
sys	0m5.624s

# mkfs -t ext4 /dev/sdb2
# time dd if=/dev/zero of=/dev/sdb2 bs=4K count=1M conv=fsync
[ext4,fsync]
4294967296 bytes (4.3 GB) copied, 74.9244 s, 57.3 MB/s

real	1m14.961s
user	0m0.244s
sys	0m9.816s

4294967296 bytes (4.3 GB) copied, 76.0565 s, 56.5 MB/s

real	1m16.091s
user	0m0.376s
sys	0m9.624s

# tux3 mkfs /dev/sdb2
# time dd if=/dev/zero of=/dev/sdb2 bs=4K count=1M conv=fsync
[WRITE_SYNC,fsync]
4294967296 bytes (4.3 GB) copied, 72.9354 s, 58.9 MB/s

real	1m12.948s
user	0m0.352s
sys	0m7.884s

4294967296 bytes (4.3 GB) copied, 72.8601 s, 58.9 MB/s

real	1m12.897s
user	0m0.428s
sys	0m7.772s

[WRITE_SYNC,FLUSH,fsync]
4294967296 bytes (4.3 GB) copied, 73.2969 s, 58.6 MB/s

real	1m13.319s
user	0m0.244s
sys	0m7.760s

4294967296 bytes (4.3 GB) copied, 72.9771 s, 58.9 MB/s

real	1m12.987s
user	0m0.408s
sys	0m7.744s

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