Jump to content


Photo

Seagate Constellation.2 and Constellation ES.2 Hard Drive Review Discu


  • You cannot start a new topic
  • Please log in to reply
17 replies to this topic

#1 TSullivan

TSullivan

    SR Admin

  • Admin
  • 688 posts

Posted 31 May 2011 - 01:52 PM

Seagate's Constellation enterprise hard drives come in 6Gbps SAS and SATA interfaces; we pit them against one another using the second-gen Constellation.2 2.5" drive and fifth-gen Constellation ES.2 3.5" drive. The following models are participating in this review:

Read Full Review

#2 continuum

continuum

    Mod

  • Mod
  • 3,581 posts

Posted 03 June 2011 - 01:32 PM

Psst, your part #s aren't quite right... Seagate ends drives with "NS" for nearline SATA, and "SS" for SAS. The 650 vs. 651 at the end indicated cache or something being different, I forget what it was...


Power consumption does look good and I see these drives are down to $330 each, which isn't bad. I wonder when the 1TB/platter nearline drives will hit (I could use a 4x1TB platter nearline...).

Also any other benchmarks of their competition? I just noticed the performance database isn't the place to go for that... although I don't think SR has reviewed the Ultrastar 7K3000 or anything so maybe not yet?

#3 Tyler

Tyler

    Member

  • Member
  • 2 posts

Posted 04 June 2011 - 12:43 AM

How can the ES.2 SAS be so much poorer performing than the SATA? Is there any word from Seagate on this? I was expecting to by about 100 of these for a new build-out, but now I'm nervous.

#4 continuum

continuum

    Mod

  • Mod
  • 3,581 posts

Posted 04 June 2011 - 02:41 AM

I think it's a firmware or a driver issue with SR's testbed that the SAS version is having. I wouldn't sweat it.

#5 Tyler

Tyler

    Member

  • Member
  • 2 posts

Posted 04 June 2011 - 12:39 PM

Hopefully they do a follow-up. I'm going to try to find other reviews on this drive to see if it was just SR that's having the issue.

#6 Kevin OBrien

Kevin OBrien

    StorageReview Editor

  • Admin
  • 1,478 posts

Posted 04 June 2011 - 04:43 PM

I am not sure it could be counted as a driver quirk on our system when the Constellation.2 SAS didn't have the same problem. If you look at the server charts the only one it maintained a similar speed to the rest of the group was the web-server which stands out being the only real-only test, with the others being a mix of R/W.

#7 fallbreak

fallbreak

Posted 06 June 2011 - 02:42 PM

Have you guys contacted Seagate or have they contacted you? Would like to hear the outcome of this discussion. Thanks!

#8 lutjens

lutjens

    Member

  • Member
  • 10 posts

Posted 04 July 2011 - 11:23 PM

Just wondering if anything further has come out of this. I was just about to buy 14 of these and I'm glad I checked out the review.
HP Proliant ML370 G6, 2x Xeon X5690, 144GB RAM, GeForce 580 GTX, 8x OCZ Vertex 3 MAX IOPS 240GB on LSi 9265-8i 1GB w/Fastpath, 12x Seagate Constellation ES.2 3TB SAS on LSi 9280-24i4e 512MB, 2x 1200W Redundant Power Supply, 2x NEC LCD3090WQXi monitors

#9 Kevin OBrien

Kevin OBrien

    StorageReview Editor

  • Admin
  • 1,478 posts

Posted 05 July 2011 - 02:53 PM

Just wondering if anything further has come out of this. I was just about to buy 14 of these and I'm glad I checked out the review.


Going to shoot a question over to them again today and see if we can get a replacement to re-test.

#10 David Szabados

David Szabados

    Member

  • Member
  • 2 posts

Posted 06 July 2011 - 12:03 PM

Going to shoot a question over to them again today and see if we can get a replacement to re-test.

Hi Kevin - I thought I'd come check out the forums here as it's been awhile. In anycase, I checked and heard back from Seagate engineering and they stated that on many of the SAS models, the prescan feature is turned on by default. This forces a read after every write and is a data-integrity checking feature. This would explain the 'flat line' experience of the 3TB model in one of the performance benchmarks. Our SAS manual talks about enabling/disabling various features through a sw utility, but we'll go ahead and send over another set so that you can test the drives with prescan off.

David Szabados
Seagate Technology

#11 TSullivan

TSullivan

    SR Admin

  • Admin
  • 688 posts

Posted 06 July 2011 - 01:55 PM

Awesome, well I guess that puts the performance issue to rest. I will give you guys an update on the results when the new drive comes in, and hopefully show the SAS model back in the range of the performance of the SATA version.

#12 lutjens

lutjens

    Member

  • Member
  • 10 posts

Posted 06 July 2011 - 10:22 PM

Hi Kevin - I thought I'd come check out the forums here as it's been awhile. In anycase, I checked and heard back from Seagate engineering and they stated that on many of the SAS models, the prescan feature is turned on by default. This forces a read after every write and is a data-integrity checking feature. This would explain the 'flat line' experience of the 3TB model in one of the performance benchmarks. Our SAS manual talks about enabling/disabling various features through a sw utility, but we'll go ahead and send over another set so that you can test the drives with prescan off.

David Szabados
Seagate Technology


Thanks for the very useful info and for taking the time to post here, David. You've laid my fears to rest regarding this issue.:)
HP Proliant ML370 G6, 2x Xeon X5690, 144GB RAM, GeForce 580 GTX, 8x OCZ Vertex 3 MAX IOPS 240GB on LSi 9265-8i 1GB w/Fastpath, 12x Seagate Constellation ES.2 3TB SAS on LSi 9280-24i4e 512MB, 2x 1200W Redundant Power Supply, 2x NEC LCD3090WQXi monitors

#13 mr_pink

mr_pink

    Member

  • Member
  • 2 posts

Posted 15 July 2011 - 11:33 AM

I've been extremely happy with the performance of the ES.2 SAS 3TB drives. Running two of them in a zfs mirror-0 config (Solaris 11 Express, attached to LSI 9211-4i storage controller via direct passthrough on VMware ESXi)
They serve, among others, as a datastore for Windows/Linux 64 bit vms.

I will run some dd and Bonnie++ benchmarks when I get the chance.

Edit: Bonnie++ 1.03 and dd benchmark results attached.

Attached Files


Edited by mr_pink, 24 July 2011 - 05:09 AM.

[Laptop] Thinkpad X300 | Ubuntu 11.04 Natty Narwhal
[Server] Supermicro X8SIL-F | Intel Xeon UP X3440 | 16GB Mushkin Proline PC3-10667E ECC | Seasonic S12II-330 | 6TB Seagate Constellation ES.2 SAS mirror-0 | SunOS sun 5.11 snv_151a

#14 afender99

afender99

    Member

  • Member
  • 4 posts

Posted 26 July 2011 - 01:09 AM

Seagate's Constellation enterprise hard drives come in 6Gbps SAS and SATA interfaces; we pit them against one another using the second-gen Constellation.2 2.5" drive and fifth-gen Constellation ES.2 3.5" drive. The following models are participating in this review:

Read Full Review


Hi,

Enjoy reading this piece of review and really appreciate your hard work on this review. I would like to check with you on some the iometer settings for the test:

1. Did you disable write cache for the iometer test?
2. For IOmeter test, some of the result you mentioned -> 2MB random transfer read/write 4k, does it means you run 2 iometer tests where the first one with 100% 2MB random read and the second one with 100% 4k random write?
3. How long is your IOMeter test duration?

Thanks!

afender99

#15 afender99

afender99

    Member

  • Member
  • 4 posts

Posted 26 July 2011 - 01:12 AM

Hi Kevin - I thought I'd come check out the forums here as it's been awhile. In anycase, I checked and heard back from Seagate engineering and they stated that on many of the SAS models, the prescan feature is turned on by default. This forces a read after every write and is a data-integrity checking feature. This would explain the 'flat line' experience of the 3TB model in one of the performance benchmarks. Our SAS manual talks about enabling/disabling various features through a sw utility, but we'll go ahead and send over another set so that you can test the drives with prescan off.

David Szabados
Seagate Technology


Hi David,

Is interesting to hear about this, is this prescan feature can be disabled by any software or can only be done in Seagate?

afender99

#16 Aman

Aman

    Member

  • Member
  • 1 posts

Posted 16 November 2011 - 03:05 PM

Great Review. Question about IOPS: In the IOMeter benchmarks with various workloads (I.e. Database 4K, Workstation 4K, etc.), we see IOPS in 100-250 range. However, in the Storage2010 HTPC Disk Capture benchmark, the IOPS for the Seagate Constellation ES.2 3TB SAS is 873. I would have expected IOPS to be lower with large IO sizes. Could you please provide some insight?

#17 TSullivan

TSullivan

    SR Admin

  • Admin
  • 688 posts

Posted 16 November 2011 - 03:11 PM

The 4K test is a pure random synthetic benchmark, so worst case fully random load. The real world traces include sequential and random mix, which accounts for higher average iops.

#18 perry2

perry2

    Member

  • Member
  • 1 posts

Posted 21 February 2012 - 06:08 AM

Hi David,

Is interesting to hear about this, is this prescan feature can be disabled by any software or can only be done in Seagate?

afender99

See http://sg.danny.cz/sg/p/sdparm.exe

Determine the disk you want to set:
sdparm -w

In the following example PD0 is assumed:

Current value of PreScan:
sdparm --get=EN_PS PD0
Enable PreScan:
sdparm --set=EN_PS=1 PD0
Disable PreScan
sdparm --set=EN_PS=0 PD0

After disable, enabl of EN_PS followed by a power cycle, the pre-scan starts. You can verify this using:
http://sourceforge.n...s/wiki/Download

smartctl -x sda
(assuming sda is your drive)

Look for line with:
Status: pre-scan is active
Number of background scans performed: 21, scan progress: 11.40%

The trouble I have is that swapping those 15k Seagate SAS disks in a system the performance drops almost 90%.

Does anyone have an idea how bad it will be disabling this pre-scan -permanently? Or any other means of preventing this performance drop?
According to the Seagate manual the performance drop should be at most 50%... and only once in a life time of the disk...


Best regards,

Perry



0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users