Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

127746

August 7th, 2013 19:00

PE T320 and S110 RAID 10 - Server 2012 - Abysmal Disk Subsystem Performance

Brand New T320 ordered with 4 500GB SATA drives (Dell/WD) set up as RAID-10.  Server 2012 Std.  32 GB RAM.  

I had good performance from a T110-II and similar 4 drive RAID-10 using the S100 controller, but the difference is night and day between the S100 and the S110.

I reinstalled several times in different configurations just to make sure it was the S110 that was the issue.  Using the latest version of CrystalDiskMark (3.02f) here are the approximate numbers I'm getting on sequential reads/writes.

 

T110-II w/S100 and 4x500GB RAID 10 -   220 / 215 MB/sec.

T320 w/S110 and 4x500GB RAID 10 -      50 / 45 MB/sec with cache configured no read ahead and write through.

T320 w/S110 and 4x500GB RAID10 -       60 / 52 MB/sec with cache configured read ahead and write back.

T320 w/S110 and 2x500GB RAID 1 -        50/45 MB/sec

T320 with S110 deactivated and just running a single 500GB drive in SATA AHCI mode.    135/133 MB/Sec.  This is about what I would expect from a single WD/Dell  7200 RPM drive.  No problem with standard SATA AHCI.

 

This is not a hardware vs software RAID issue.  Like I mentioned, I get great numbers on a T110-II with S100 in RAID 10.

Dell - I can't believe you ship something this bad.  My preference is to figure out why it's so bad and fix the issue, but if I don't get answers soon the server will be going back to Dell.  The server as it sits now is unusable.  I've got 10 year old servers sitting in my salvage pile that perform better than this.

[ADMIN NOTE: Profanity removed per TOU] Dell ???

 

26 Posts

August 8th, 2013 09:00

Info on CrystalDisk Mark can be found here:

http://crystalmark.info/software/CrystalDiskMark/manual-en/MainWindow.html

the numbers are sequential 512k/ qd1 for the values that I listed in my initial message.

 

What got me started on this was the new server felt extremely sluggish during the install and config.  So I tried a couple of disk benchmarks and they all showed extremely poor performance.  I listed CrystalDiskMark numbers because it is widely known and used in benchmarking. 

 

7 Technologist

 • 

16.3K Posts

August 7th, 2013 22:00

Open a case with Dell Support ... you can't do that here.

August 8th, 2013 04:00

Hi,

 

For these numbers can you quote the IO sizes & Q depths running @ the time when you took the readings?

-Deepu

 

 

October 21st, 2014 15:00

Sorry to bring up such an old thread, but were you ever able to figure out a solution to this problem?  We are seeing slow performance on R220 servers with S110 controllers running RAID 1, but have no issue with our T110's.  Latest drivers/firmware/bios do not seem to be of benefit.  

December 24th, 2015 09:00

I know this is old, but I am desperate.  What ended up being the resolution to this? I am having this exact same issue today and Dell is telling me it is a hardware vs. software RAID issue, and I know it is not.

26 Posts

December 24th, 2015 14:00

If I recall correctly I just had to wait for the background array initialization to finally complete (it may take several hours or more) and then the performance improved a bit.  Download Dell OMSA (Open Manager Server Administrator) and install it on the server in question.  It will allow you to monitor the progress of the Background Init process on the drive array.  Once the init process is complete, you should see performance improve.

www.dell.com/.../OMSA

FWIW, I just jumped onto the server that I started this thread about and ran a Crystal Diskmark again and the numbers are as follows:

4 x 500 GB SATA drives in RAID-10

-----------------------------------------------------------------------

CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo

                          Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

          Sequential Read :   189.667 MB/s

         Sequential Write :   127.642 MB/s

        Random Read 512KB :    36.863 MB/s

       Random Write 512KB :    92.901 MB/s

   Random Read 4KB (QD=1) :     0.569 MB/s [   139.0 IOPS]

  Random Write 4KB (QD=1) :     2.929 MB/s [   715.1 IOPS]

  Random Read 4KB (QD=32) :     2.898 MB/s [   707.5 IOPS]

 Random Write 4KB (QD=32) :     2.939 MB/s [   717.4 IOPS]

 Test : 1000 MB [D: 11.9% (99.2/832.8 GB)] (x5)

 Date : 2015/12/24 13:43:00

   OS : Windows Server 2012 Server Standard Edition (full installation) [6.2 Build 9200] (x64)

IMHO, the Dell S110 controllers are rather poor performers.  They were barely adequate in performance when they were introduced and since then mechanical drives are faster and SSDs are prevalent.  Sadly the S110 and associated are still being offered by Dell and the performance of these controllers is very sub par.

I now use the Dell T20 for many of my smaller Server builds.  It is not saddled with the S110 controller and instead is a pure Intel RST solution.  I find the Intel RST performance is about as good as it gets without a dedicated CPU for the RAID controller.

Here are some numbers for the T20 in RAID-10 with 4 Seagate 1TB Constellation drives and Intel RST:

-----------------------------------------------------------------------

CrystalDiskMark 3.0.4 x64 (C) 2007-2015 hiyohiyo

                          Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

          Sequential Read :   309.817 MB/s

         Sequential Write :   295.623 MB/s

        Random Read 512KB :    44.745 MB/s

       Random Write 512KB :   114.239 MB/s

   Random Read 4KB (QD=1) :     0.600 MB/s [   146.5 IOPS]

  Random Write 4KB (QD=1) :     2.312 MB/s [   564.5 IOPS]

  Random Read 4KB (QD=32) :     4.226 MB/s [  1031.7 IOPS]

 Random Write 4KB (QD=32) :     2.405 MB/s [   587.1 IOPS]

 Test : 1000 MB [D: 28.5% (977.0/3433.0 GB)] (x5)

 Date : 2015/12/24 14:05:01

   OS : Windows Server 2012 R2  [6.3 Build 9600] (x64)

and just for comparisons sake, a T20 with two 1TB Crucial SSD's in RAID-1 and Intel RST:

-----------------------------------------------------------------------

CrystalDiskMark 3.0.3 x64 (C) 2007-2013 hiyohiyo

                          Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

          Sequential Read :  1043.186 MB/s

         Sequential Write :   487.407 MB/s

        Random Read 512KB :   805.658 MB/s

       Random Write 512KB :   411.275 MB/s

   Random Read 4KB (QD=1) :    31.286 MB/s [  7638.1 IOPS]

  Random Write 4KB (QD=1) :    94.559 MB/s [ 23085.6 IOPS]

  Random Read 4KB (QD=32) :   549.455 MB/s [134144.3 IOPS]

 Random Write 4KB (QD=32) :   264.189 MB/s [ 64499.3 IOPS]

 Test : 1000 MB [D: 11.5% (100.7/875.7 GB)] (x5)

 Date : 2015/12/24 14:19:55

   OS : Windows Server 2012 R2 Server Standard (full installation) [6.3 Build 9600] (x64)

As you can see, the Dell T20 with SSD's and Intel RST make a great combo.  Equip the Xeon T20 with 32 GB of RAM and SSD's and you've got a great system for running several VM's.

4 Operator

 • 

1.8K Posts

December 26th, 2015 07:00

Do you have a number for the CPU usage by the Intel RST during heavy usage?

26 Posts

December 26th, 2015 09:00

With 100% mechanical drives the CPU usage is 6% or less, even with QD=32.  

With SSD's it gets a bit higher, but you're moving substantially more data.  Sequential is about 10% for the RAID-1 pair, QD=32 can be as high as 30% but it only seems to use 1-2 threads.  With a 4 or 8 thread processor it does not seem to impact performance at all.  In systems where I have configured a RAID-1 SSD pair along with another RAID-1 mechanical pair I'm able to run Crystal DiskMark simultaneously testing each pair and the numbers stay pretty consistent.  

If you look at the numbers in my last post, comparing QD=32 IOPs, Mechanical vs SSD, the SSD numbers are over 100x higher.  I'd expect to use a bit more of the CPU to move 100x more data.

And please keep in mind that these numbers are from a synthetic benchmark.  In the real world I've had no trouble running 4 or 5 VM's on one of these T20's, equipped with a Xeon CPU, 32 GB RAM, 1 TB SSD's in RAID-1 along with 2 TB SATA Mechanicals also in RAID-1.  

I hope this information is helpful.

December 28th, 2015 07:00

Thanks for the reply. Our server takes a full ten minutes from power on until crtl-alt-del prompt and the hard drives are constantly being hit, even with nobody on the system. I have uninstalled everything, turned off all indexing and shadow copies and the drives are still constantly active. Dell swapped the motherboard and cable to no affect, so I am not sure what we are going to do next. I am in the same situation. I have a 10 year old clone server in my scrap heap that is faster than this thing.

January 4th, 2016 10:00

My experience with this issue was way more than just poor performance.  As Andy mentioned, my servers were completely unusable - like running a machine pegged at 100% CPU (but hardly any CPU actually being used).  I haven't looked at this since I posted.  We ended up forcing Dell to replace the servers with ones that included the low end hardware raid controller option.  They run fine.  I hoped/assumed it was a driver issue that Dell would have solved by now (been over a year).  Unfortunately it looks like Dell has not released a newer driver for the S110 controller for Server 2012.  If I remember right, Server 2008R2 ran fine - it was just Server 2012 that was horrible.  Maybe try the latest BIOS (Dec 2015)?  

No Events found!

Top