1

I have a new PowerEdge T320. Details are:

  • Intel Xeon E5-1410 @ 2.8 GHz
  • 16GB
  • Windows Server 2008 r2 configured as Domain Controller
  • Dell PERC S110 Software Raid
  • 4x 7200 rpm 1TB drives configured in Raid5

Clients initially started complaining about Quickbooks getting disconnected. In troubleshooting, I found that I am getting what I see as a terribly slow read speed on my hard drive.

HDTune, Dskspeed and ATTO all show about 5MB/s read speed. ATTO showed up to 15MB/s when reading 1024 blocks.

I have one Dell support agent telling me "you shouldn't use software raid" and another telling me, "this is not that bad".

So, my questions:

  1. Does a read speed that slow indicate a problem with one of my drives, or is it to be expected with software raid?

  2. If it is a symptom of a problem, is it possible that it would cause an application like Quickbooks to fail pulling data from server and, as a result, stop responding.

  3. Is there a way to pull out a drive from server, test the read write speed on another machine, and reconnect it without rebuilding the drive?

4
  • 1
    Sounds ridiculous to me. You could call the DELL sales people and ask them what performance you can expect in that configuration, ask them if there's a lower limit they would guarantee. ;-) I had a look at Google and all I find are statements about several hundreds of MiB/s, even in software. Feb 23, 2013 at 18:02
  • I suspect such speed is caused by degraded RAID, which means one HDD is failed. Maybe you can check the S.M.A.R.T status of your HDDs. I use linux software RAID5, when I copying millions of audio files to a degraded RAID5, it's speed never reached 10MB/s. Feb 23, 2013 at 18:03
  • Raid controller is telling me all good, and dell ran built in diags. I don't know if diags check smart, but I know raid controller doesn't and standalone app from WD can't see thru raid configuration. So, it appears that all drives are good. And, beacause of this problem, it takes days to rebuild drives. So, im very hesitant to pull them out, one by one, for testing and rebuilding. Especially since server has mission critical data, rebuilding will further decrease performance, and any backup will have missing data a day or two later when rebuilding is almost done and a drive crashes. Feb 23, 2013 at 18:50
  • I will post comment once problem is resolved. Dell is currently sending me a new hard drive to test. Feb 26, 2013 at 16:46

2 Answers 2

4

That is absolutely not acceptable performance for a RAID setup, software or hardware. An individual SATA drive would outperform that by many times. Could you provide more detail around what kind of drives you are using? Are you sure your array isn't already running with a failed disk, in a degraded state? I also know that using "green" SATA hard drives can cause problems in RAID arrays due to their built-in power management.

To answer your symptom of a problem question, check the System event log for disk and file system errors. Failures to read or write in a timely enough fashion will be logged.

You can pull a single drive at a time out of a RAID-5, but while the disk is removed your array is performing in a degraded state, and you will have to replace the drive and rebuild completely before removing another drive. If you remove a second drive before the first drive rebuilds, you will cause the array to fail. Be careful not to do this with a live disk if you already have a failed disk, as I referred to previously as a possibility.

I have an LSI 8308 ELP SAS/SATA hardware RAID controller in a lab box with 4 old 250 GB SATA-II drives installed on it in a RAID-5, and it reads/writes at around 300 MB/sec, and I have 2 SATA-II SSDs on my workstation on a software RAID (ICH10R) that reads/writes 500 MB/sec. In my case, these are both either very legacy or consumer-class devices that are outperforming your server by 60-100x.

4
  • Thanks for the response. I am using 4 "Western Digital WD RE4 WD1003FBYX 1TB 7200 RPM 64MB Cache SATA 3.0Gb/s 3.5". I don't believe these have any power management built-in. I will check system event log now. Feb 23, 2013 at 18:13
  • I see nothing in logs. All standard stuff and a few unrelated errors. Feb 23, 2013 at 18:51
  • Can you see anything useful in the OpenManage logs? If this is a brand new box, is the controller doing any kind of background initialization? Feb 24, 2013 at 17:05
  • Pulled all drives to test via a sata-to-usb cable. All read/write was around 40 Mbps, which is cap with my adapter. Nonetheless, everything tested fine. Did not see anything in OpenManage. Everything looked perfectly fine in status of controller as well. I did notice drives are set to read ahead/write through, and I found an old data sheet about there being a problem with that setting and the intel C600 chipset. Of course, since it is a dell raid controller, I don't know that the intel chipset has anything to do with it. Any more suggestions? Mar 2, 2013 at 16:32
0

Had the same issue with almost the same setup. The Perc S110 software raid is horrible and when you are throwing in parity it is even worse. To Fix this issue i recommend changing it to a Raid 10. It Flies after the change but you will have to back everything up...

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .