1

I have a Dell PowerEdge R820 server which is under maintenance by other third party. There are 6 SAS (10K RPM, 6gbps) disks and they are configured as RAID 5 using PERC controller.

Currently I am facing performance issue with the server. Basically it is with the disk. When I tried to write 4GB of data, it is taking 12 minutes to complete. I am using a Linux Server.

Please see the output of dd command:

# # time dd if=/dev/zero of=TestFile bs=4096 count=1024000
1024000+0 records in
1024000+0 records out
real    12m 3.56s
user    0m 7.94s
sys 0m 0.00s

I have also checked with the other Desktop made server, where RAID 5 is configured with 4 SATA (7.2K RPM) disks. It is taking only 19 seconds to write 4GB of data to the disk.

I can see the clear problem of disk I/O performance issue. But the third party is denying, they are telling that, this is the normal time. But I refuse to agree with them.

Can you please tell me what should be the normal time to write 4GB data to the volume configured with 6SAS (10K RPM) disks?

closed as off-topic by sourcejedi, Archemar, MelBurslan, Scott, Anthon Aug 25 '16 at 4:25

  • This question does not appear to be about Unix or Linux within the scope defined in the help center.
If this question can be reworded to fit the rules in the help center, please edit the question.

  • The file system matters though. You should also consider using oflag=direct or oflag=nocache,sync for a realistic write speed. A faulty drive might also slow down the raid performance. – Valentin Bajrami Aug 24 '16 at 8:25
  • 4
    I'm voting to close this question as off-topic because I think it should be moved to ServerFault. – sourcejedi Aug 24 '16 at 8:38
2

That does seem like a disk performance issue.

You should get something in between 20 MB/s to 80 MB/s depending on block size I think. I found this old 10k disk comparison where you can see how different drives are performing http://techreport.com/review/5236/10k-rpm-hard-drive-comparison/7 .

I also found a thread from dell forum where someone is facing same kind of issue: http://en.community.dell.com/support-forums/servers/f/906/t/19475037

To answer you question: No 5-6 MB/s is not normal.

  • JFC. That's ugly engineering. (Sounds as if it uses some high-overhead method to close the RAID5 write hole w/o having any separate non-volatile memory. I.e. a data journal, requiring synchronous writes). – sourcejedi Aug 24 '16 at 8:34
1

That's a reasonable test for sequential writes. General hard disk performance should be in the region of 100MB/s. You should include a final sync in the timings, which could be why your desktop appears to run 2x that speed.

You might see variance around +-30 on different types of drives. RAID can affect that, but in general it shouldn't cause a significant slowdown, not to sequential writes. (Maybe if you had used conv=sync).

Your server should not be running sequential writes at 5MB/s, that is just wrong. Note that one cause of slow performance is a failing disk.

Not the answer you're looking for? Browse other questions tagged or ask your own question.