Some disk write performance statistics on a Raid-50

Written by Claudio Kuenzler - 0 comments

Published on May 26th 2014 - Listed in Linux Hardware

---

Last week I installed a new server (HP Proliant DL 380 Gen8) with Ubuntu 14.04 LTS. The server has 8 SAS disks, all attached to a P420i storage controller with a 1GB cache module.

The 8 disks are bundled together in a RAID-50. The performance testing bounced me off the chair!

Write 1.1 GB file

time { dd if=/dev/zero of=file.tmp bs=32768 count=32768; sync; }
32768+0 records in
32768+0 records out
1073741824 bytes (1.1 GB) copied, 1.65159 s, 650 MB/s

real    0m2.228s
user    0m0.068s
sys     0m1.594s

Note the big block size... So that's not very realistic. To do a test closer to reality, lets use the same block size as the actual file system:

blockdev --getbsz /dev/sda1
4096

Write 4.2 MB file

time { dd if=/dev/zero of=file.tmp bs=4096 count=1024; sync; }
1024+0 records in
1024+0 records out
4194304 bytes (4.2 MB) copied, 0.00492218 s, 852 MB/s

real    0m0.173s
user    0m0.002s
sys     0m0.169s

Write 4.2 GB file

time { dd if=/dev/zero of=file.tmp bs=4096 count=1024000; sync; }
1024000+0 records in
1024000+0 records out
4194304000 bytes (4.2 GB) copied, 4.39035 s, 955 MB/s

real    0m6.993s
user    0m0.783s
sys     0m3.612s

Write 42 GB file

time { dd if=/dev/zero of=/mnt/file.tmp bs=4096 count=10240000; sync; }
10240000+0 records in
10240000+0 records out
41943040000 bytes (42 GB) copied, 55.0411 s, 762 MB/s

real    0m57.426s
user    0m8.667s
sys     0m42.449s

Sweeeeeeeeet.

Add a comment

Show form to leave a comment

Comments (newest first)

No comments yet.