THE SQL Server Blog Spot on the Web

Welcome to SQLblog.com - The SQL Server blog spot on the web Sign in | |
in Search

Joe Chang

HP Storage Works 2000 sa and fc G2 performance specs

While reading through the documentation for the HP Storage Works 2000 MSA, I found the following performance numbers cited in the quickspecs for both the FC and SAS versions. The 2000fc version cited performance numbers for the fc (Fiber Channel), sa (SAS) and i (iSCSI) models. The 2000sa version only cites SAS results. The MSA 2000 Technical Cookbook also cites performance results.

The results cited in the FC model documentation for the SAS and iSCSI models cannot be correct, unless the SAS and iSCSI models were completely botched, and as far as I am aware of, it has been a really long time since HP/Compaq last botched a design this badly.

I am inclined to think that the results cited in the FC section are incorrect, and may inadvertently be used by HP sales rep to recommend the FC version. The configuration in the performance report below is somewhat ok for OLTP, 4 x 4Gbs FC ports or equivalent. Of course if I had 4 SAS RAID controllers, each connected to 2 MSA-60 with 12 15K HDDs, I could get 4-6GB/s sequential. For a clustered transaction processing system, you pretty much need the SAN to support shared disks. But I am inclined to ask people to consider database mirroring, each node employing direct attach storage.

For Data Warehouse applications, I would outright recommend the direct-attach MSA-60 and skip the SAN.

 

From HP StorageWorks 2000fc G2 Modular Smart Array on 2009 Sep 14

Performance numbers are a guideline as established by tests using RAW I/O in an Operating System Agnostic test lab environment. A total of 96 72GB 15K SAS drives were used in a dual controller configuration of eight 12 Disk vdisks, eight 806 GB volumes, and two volumes per host. Four hosts directly attached to the MSA2000fc G2 were used in this test configuration (results cannot be expected with a single host). Results were achieved in Sequential Writes with 256K blocks; all random tests were based on 8K block sizes.
Number and type of applications, drive type, operating system used, and the number of hosts will affect overall performance.

 

Workload MSA2300fc MSA2000sa MSA2000i
Host Connect 4 Gb Fibre Channel 3 Gb SAS 1GbE Ethernet
       
MSA2000 RAID 10 Performance Results (rounded)
Random Reads IOPs 21,400 10,600 8,200
Random Writes IOPs 8,800 4,900 4,500
Random Mix IOPs 60/40 read/write 13,500 6,800 6,100
Sequential Reads MB 1,300 700 300
Sequential Writes MBs 560 350 260
MSA2000 RAID 5 Performance Results (rounded)
Random Reads IOPs 20,500 10,200 7,800
Random Writes IOPs 2,400 2,000 1,600
Sequential Reads MBs 1,300 700 300
Sequential Writes MBs 780 380 270
Random Mix IOPs
60/40 read/write
5,900 3,300 3,200
MSA2000 RAID 6 Performance Results (rounded)
Random Reads IOPs 20,500 10,100 7,800
Random Writes IOPs 1,600 1,400 1,200
Sequential Reads MBs 1,300 700 300
Sequential Writes MBs 820 380 270
Random Mix IOPs 60/40 read/write 4,500 2,750 2,560

 

From: HP StorageWorks 2000sa G2 Modular Smart Array

Performance numbers are a guideline as established by tests using RAW I/O in an Operating System Agnostic test lab environment. A total of 96 72GB 15K SAS drives were used in a dual controller configuration of eight 12 Disk vdisks, eight 806 GB volumes, and two volumes per host. Four hosts directly attached to the MSA2000sa G2 were used in this test configuration (results cannot be expected with a single host). Results were achieved in Sequential Writes with 256K blocks; all random test were based on 8K block sizes.
Number and type of applications, drive type, operating system used, and the number of hosts will affect overall performance.

 

Workload
MSA2300sa G2
Host Connect 3Gb SAS
MSA2000 RAID 10 Performance Results (rounded)
Random Reads IOPs
21861
Random Writes IOPs
14491
Random Mix IOPs 60/40 read/write
17700
Sequential Reads MB
1049
Sequential Writes MBs
531
MSA2000 RAID 5 Performance Results (rounded)
Random Reads IOPs
21019
Random Writes IOPs
2705
Sequential Reads MBs
1051
Sequential Writes MBs
644
Random Mix IOPs
60/40 read/write
5897
MSA2000 RAID 6 Performance Results (rounded)
Random Reads IOPs
21029
Random Writes IOPs
1918
Sequential Reads MBs
1054
Sequential Writes MBs
736
Random Mix IOPs 60/40 read/write
4872

The following slide is from the HP MSA2000 Technical Cook Book.

MSA 2000 Technical Cookbook

 

Update 2009 Oct 30

Apparently HP has updated the HP StorageWorks 2000fc G2 Modular Smart Array spec sheet

Performance numbers are a guideline as established by tests using RAW I/O in an Operating System Agnostic test lab environment. A total of 96 72GB 15K SAS drives were used in a dual controller configuration of eight 12 Disk vdisks, eight 806 GB volumes, and two volumes per host. Four hosts directly attached to the MSA2000fc G2 were used in this test configuration (results cannot be expected with a single host). Results were achieved in Sequential Writes with 256K blocks; all random tests were based on 8K block sizes.
NOTE: Number and type of applications, drive type and number of drives, operating system used, and the number of hosts will affect overall performance.

 

MSA2 G2 Performance MSA2300fc G2 MSA2000sa G2 MSA2000i G2
Protocol (host connect) 4 Gb
Fibre Channel
3 Gb
SAS
1GbE Ethernet
MSA2000 RAID 10 Performance Results
Random Reads IOPs 22,874 21,861 13,658
Random Writes IOPs 15,008 14,491 14,399
Random Mix IOPs 60/40 read/write 17,878 17,700 13,493
Sequential Reads MB 1,238 1,050 274
Sequential Writes MBs 532 532 266
MSA2000 RAID 5 Performance Results
Random Reads IOPs 22,044 21,000 12,335
Random Writes IOPs 2,714 2,714 2,714
Random Mix IOPs 60/40 read/write 5,325 4,926 5,325
Sequential Reads MBs 1,238 1,050 274
Sequential Writes MBs 725 617 266
MSA2000 RAID 6 Performance Results
Random Reads IOPs 21,975 21,000 12,292
Random Writes IOPs 1,876 1,876 1,876
Random Mix IOPs 60/40 read/write 4,293 3,865 3,851
Sequential Reads MBs 1,238 1,050 274
Sequential Writes MBs 772 729 266

 

Published Monday, September 14, 2009 3:38 PM by jchang
Filed under: ,

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS

Comments

 

StorageMonkey said:

Are you saying to got "4-6GB/s sequential" performance based on testing you did, or that you *should* get that (in theory) based on the numbers HP has provided on the MSA2000fc?  I'm looking to test 4 clustered MSA2000fc G2's and am hoping to hit around 4GB's/sec for sequential reads (with 192 72GB 15k drives) and was wondering if you already tested similar numbers?  

September 29, 2009 4:11 AM
 

jchang said:

The 4 SAS RAID controllers each connected to 2 MSA60 each with 12 15K disk is a projection.

Of course, the ProLiant DL785G5 has already measured 7GB/s over 6 P800 controlllers from a SQL Server table scan. (close 9GB/s from a pure IO test, something like IOmeter).

I am projecting down from a measured result, not up based on what a single controller can do and .

My last measurement on system with a good direct-attach storage did 3.2GB over 4 SAS RAID controller, or 800MB/s per adapter.

This was a Xeon 7400 generation system, PCI-E gen 1 (2.5Gbps) and SAS 3Gbps. Two of the 4 adapters were in x8 slots and 2 in x4.

The system had 4 x8 slots, but each pair shared one x8 uplink bandwidth.

The reason I am projecting for 4 controllers is that of the Xeon 5400 series systems (on the 5000P/X chipset) and the Xeon 7400 series with the 7300 chipset, no systems had 4 x8 PCI-E slots (with dedicated bandwidth). Intel did have 5400 chipset for the 5400 series, but the major vendors did not have system based on it. SuperMicro did, but was not inclined to build another system from scratch.

Now the 2-way Dell PowerEdge T710 does have 4 x8 PCI-E Gen 2 slots. The HP ProLiant DL370G6 has 2 x16, 2x8 (1 used for NICs) and 6 x4 slots, also PCI-E Gen2.

If you have a PCI-E Gen 2 SAS RAID controller, you should be able to get 1.6GB/sec on a x4 slot.

Of course, I cannot afford 8 MSA 60, each with 12 drives (about $56K?) so I stated the 4-6GB/sec target to get some one to buy the equipment.

So, will you get 4GB/s on 4 MSA2000fc with 192 drives?

The HP data says one pair of MSA2000fc with 96 drives could do 1300MB/s tested with 4 hosts. It also states "results cannot be expected with a single host" but I am inclined to disagree because I have set up single host with heavy IO many times. And so do all the other benchmark people.

So do you have 4 MSA2000fc controller pairs, or 2 pairs?

Expect only 2600MB/s from 2 pairs.

In direct attach, I can get the full disk sequential transfer rate from a single disk (125-150MB/s) and I can saturate the x4 SAS channel with 12 disks.

In a SAN, I typically see about 10MB/sec per disk. Meaning 192 disks should do 1900MB/sec. My understanding is 20-30MB/sec per disk is possible, but requires tuning.

September 29, 2009 10:57 AM
 

StorageMonkey said:

Thanks for the info!  I will be looking to test 4 MSA2000fc controller pairs.  I will also have 4 DL580's that will be SAN connected to the MSA's (via a pair of 8GB 24 port switches, although you've got me thinking about direct-connect based on your comments).  I'll have to go back and verify if the DL580's have the PCI-E Gen 2 SAS RAID controller.

September 29, 2009 12:15 PM
 

Luke Youngblood said:

I recently benchmarked 4 MSA2000fc G2 arrays together with an Oracle RAC data warehouse/DSS workload.  If you would like to see the results of these tests, please visit my blog at the address above.

March 10, 2010 10:10 AM
 

Logicalis HP Storage said:

I'm having trouble setting up my benchmark. Can anyone help?

April 21, 2010 3:22 PM
 

jchang said:

You are having trouble? I just tried doing the Microsoft Fast Track Data Warehouse with the MSA 2000 G2 fc version. We had huge difficulties getting it setup. Turns out there was a partially dead (LED?) module in the switch and yet the MSA reported seeing the port/host. The FTDW document did not mention installing the QLogic SanSurfer or Emulex HBAnywhere software which would tell you all this quickly. Also, the Microsoft Windows Server 2008 MPIO module was a pain, had to unistall and reinstall. Perhaps Luke Youngblood can help, but you might contact him directly.

April 21, 2010 4:24 PM

Leave a Comment

(required) 
(required) 
Submit

About jchang

Reverse engineering the SQL Server Cost Based Optimizer (Query Optimizer), NUMA System Architecture, performance tools developer - SQL ExecStats, mucking with the data distribution statistics histogram - decoding STATS_STREAM, Parallel Execution plans, microprocessors, SSD, HDD, SAN, storage performance, performance modeling and prediction, database architecture, SQL Server engine

This Blog

Syndication

Powered by Community Server (Commercial Edition), by Telligent Systems
  Privacy Statement