Disk Performance Hands On, Part 1: Test Harness and Results

<div><p><em>This is Part 1 in an ongoing series on disk performance. You can read the entire series by starting at the <a href="http://kendalvandyke.blogspot.com/2009/02/disk-performance-hands-on-series.html">Introduction</a>.</em></p> <p>Here are the specs on the hardware I used to do my tests and how I went about my testing:</p> <p><strong>Test Harness 1: Dell PowerEdge 2950</strong></p> <ul> <li>Dell PowerEdge 2950 </li> <li>Windows Server 2003 R2 Standard Edition Service Pack 2 patched with all updates available at test time </li> <li>2 x Dual Core Xeon 5130 @ 2GHz (2 cores per CPU = 4 cores total) </li> <li>4 GB RAM </li> <li>PERC 5/i integrated controller </li> <li>Disk cache policies set to write back and no read ahead </li> <li>73 GB 15,000 RPM Seagate Cheetah SAS hard drives </li> <li>RAID configurations: <ul> <li>RAID 1: Physical Disks 0:0:2 and 1:0:4 <ul> <li>Allocation unit size (a.k.a. block size): 64KB </li> <li>Partition offset: 32KB, 64KB, 1024KB </li> </ul> </li> <li>RAID 5: Physical Disks 0:0:2, 0:0:3, 1:0:4, and 1:0:5 <ul> <li>RAID Stripe Size: 8KB, 64KB, 128KB </li> <li>Allocation unit size (a.k.a. block size): 4KB, 8KB, 64KB </li> <li>Partition offset: 32KB, 64KB </li> </ul> </li> <li>RAID 10: Physical Disks 0:0:2 and 0:0:3 (Span 0); 1:0:4 and 1:0:5 (Span 1) <ul> <li>RAID Stripe Size: 8KB, 64KB, 128KB </li> <li>Allocation unit size (a.k.a. block size): 4KB, 8KB, 64KB </li> <li>Partition offset: 32KB, 64KB </li> </ul> </li> </ul> </li> <li>SQLIO v1.5.SG <ul> <li>File size: 64 GB </li> <li>Threads: <ul> <li>RAID 1: 4 </li> <li>RAID 10 & 5: 8 </li> </ul> </li> <li>Thread depth: 8 </li> <li>Buffering: Hardware </li> <li>Test duration: 120 seconds </li> <li>Wait between tests: 60 seconds </li> </ul> </li> </ul> <p><strong>Test Harness 2: Dell PowerEdge 2950 w\ PowerVault MD1000</strong></p> <ul> <li>Dell PowerEdge 2950 </li> <li>Windows Server 2003 R2 Standard Edition Service Pack 2 (x64) patched with all updates available at test time </li> <li>2 x Quad Core Xeon E5420 @ 2.50GHz (4 cores per CPU = 8 cores total) </li> <li>4 GB RAM </li> <li>PERC 5/E controller connected to a PowerVault MD1000 drive enclosure operating in split bus mode </li> <li>Disk cache policies set to write back and no read ahead </li> <li>146 GB 15,000 RPM Seagate Cheetah SAS hard drives </li> <li>RAID configurations: <ul> <li>RAID 1: Physical Disks 0:0:7 and 1:0:0 </li> <li>RAID 5: Physical Disks 0:0:7, 0:0:8, 0:0:9, 0:0:10, 0:0:11, 0:0:12, 1:0:0, 1:0:1, 1:0:2, 1:0:3, 1:0:4, and 1:0:5 </li> <li>RAID 10: Physical Disks 0:0:7 and 1:0:0 (Span 0); 0:0:8 and 1:0:1 (Span 1); 0:0:9 and 1:0:2 (Span 2) ; 0:0:10 and 1:0:3 (Span 3); 0:0:11 and 1:0:4 (Span 4); 0:0:12 and 1:0:5 (Span 5) </li> </ul> </li> <li>RAID stripe size: 64KB </li> <li>Allocation unit size (a.k.a. block size): 64KB </li> <li>Partition offset: 32KB, 64KB, 1024KB </li> <li>SQLIO v1.5.SG <ul> <li>File size: <ul> <li>RAID 1: 128 GB </li> <li>RAID 10: 512 GB </li> <li>RAID 5: 512 GB </li> </ul> </li> <li>Threads: 4, 8, 16 </li> <li>Thread depth: 8 </li> <li>Buffering: Hardware </li> <li>Test duration: 120 seconds </li> <li>Wait between tests: 60 seconds </li> </ul> </li> </ul> <p><strong></strong></p> <p><strong></strong></p> <p><strong>Test Harness 3: Dell PowerEdge 2850 w\ PowerVault 220s</strong></p> <ul> <li>Dell PowerEdge 2850 </li> <li>Windows Server 2003 R2 Standard Edition Service Pack 2 patched with all updates available at test time </li> <li>2 x Dual Core Xeon @ 3.20GHz (2 cores per CPU = 4 cores total) </li> <li>4 GB RAM </li> <li>PERC 4/DC controller connected to a PowerVault 220s drive enclosure operating in split bus mode </li> <li>Disk cache policies set to write back and no read ahead </li> <li>146 GB 15,000 RPM Seagate Cheetah Ultra320 SCSI hard drives </li> <li>RAID configurations: <ul> <li>RAID 5: Physical Disks 0:0, 0:1, 0:2, 0:3, 0:4, 0:5, 1:9, 1:10, 1:11, 1:12, 1:13, and 1:14 </li> <li>RAID 10: Physical Disks 0:0 and 1:9 (Span 0); 0:1 and 1:10 (Span 1); 0:2 and 1:11 (Span 2) ; 0:3 and 1:12 (Span 3); 0:4 and 1:13 (Span 4); 0:5 and 1:14 (Span 5) </li> </ul> </li> <li>RAID stripe size: 64KB </li> <li>Allocation unit size (a.k.a. block size): 64KB </li> <li>Partition offset: 32KB, 64KB, 1024KB </li> <li>SQLIO v1.5.SG <ul> <li>File size: <ul> <li>RAID 10: 512 GB </li> <li>RAID 5: 512 GB </li> </ul> </li> <li>Threads: 4, 8, 16 </li> <li>Thread depth: 8 </li> <li>Buffering: Hardware </li> <li>Test duration: 120 seconds </li> <li>Wait between tests: 60 seconds  </li> </ul> </li> </ul> <p><strong></strong></p> <p><strong>Results <br /></strong>Here are the results of every test that I ran so that you can analyze all the nitty-gritty numbers for yourself. (Note that these are formatted for Excel 2007. If you don’t have Excel 2007 you can still view them by downloading the Excel Viewer from <a href="http://www.microsoft.com/downloadS/details.aspx?familyid=1CD6ACF9-CE06-4E1C-8DCF-F33F669DBC3A&displaylang=en" target="_blank">here</a>)</p> <ul> <li><a href="http://kendal-van-dyke.googlegroups.com/web/Disk+Performance+Hands+On+-+PowerEdge+2950.xlsx?gda=ALifnWEAAAB7AxIJCM8IJwyCDPmKr7w4Bz7Qq__tmbVDponPcXvlsBqm2L6n5nIHaEinlScw5LmFIO3DHjvsgTTlN6hiRGEC3DqFm06UqdQE_m0nH8EE01MJP-kItwVPBwjdKujmMDOVcV4Kf5x1iV4X6-2IalYA&gsc=U3dtShYAAABq66eG9pHZMyvDs9hsIwQeg-kXU5InE09W2o0GCSVgCQ">Disk Performance Hands On - PowerEdge 2950.xlsx</a> </li> <li><a href="http://kendal-van-dyke.googlegroups.com/web/Disk+Performance+Hands+On+-+PowerVault+220S.xlsx?gda=Lt3FfGIAAAB7AxIJCM8IJwyCDPmKr7w4Bz7Qq__tmbVDponPcXvlsBqm2L6n5nIHaEinlScw5LmyhQAnbsmPXf8jG234XMu1splxFIEoGffjGfetcsl3oVXi7dpriIAjJhAipsb2do-CHqjxxwsG8_oKG53kozMh&gsc=U3dtShYAAABq66eG9pHZMyvDs9hsIwQeg-kXU5InE09W2o0GCSVgCQ">Disk Performance Hands On - PowerVault 220S.xlsx</a> </li> <li><a href="http://kendal-van-dyke.googlegroups.com/web/Disk+Performance+Hands+On+-+PowerVault+MD1000.xlsx?gda=WcwzbWQAAAB7AxIJCM8IJwyCDPmKr7w4Bz7Qq__tmbVDponPcXvlsBqm2L6n5nIHaEinlScw5LmyhQAnbsmPXf8jG234XMu1K6RA6foUu1cql6jylTsmkFXq71KIRN2DRDZ98DIdT53NzgFmQudIVZfn2evkHEao&gsc=U3dtShYAAABq66eG9pHZMyvDs9hsIwQeg-kXU5InE09W2o0GCSVgCQ">Disk Performance Hands On - PowerVault MD1000.xlsx</a> </li> </ul> <p>Stay tuned…In <a href="http://kendalvandyke.blogspot.com/2009/02/disk-performance-hands-on-part-2-raid.html">Part 2</a> I will take a look at how partition offset, RAID stripe size, and allocation unit size affect RAID 10 performance.</p> </div>

This is Part 1 in an ongoing series on disk performance. You can read the entire series by starting at the Introduction.

Here are the specs on the hardware I used to do my tests and how I went about my testing:

Test Harness 1: Dell PowerEdge 2950

• Dell PowerEdge 2950
• Windows Server 2003 R2 Standard Edition Service Pack 2 patched with all updates available at test time
• 2 x Dual Core Xeon 5130 @ 2GHz (2 cores per CPU = 4 cores total)
• 4 GB RAM
• PERC 5/i integrated controller
• Disk cache policies set to write back and no read ahead
• 73 GB 15,000 RPM Seagate Cheetah SAS hard drives
• RAID configurations:
  • RAID 1: Physical Disks 0:0:2 and 1:0:4
    • Allocation unit size (a.k.a. block size): 64KB
    • Partition offset: 32KB, 64KB, 1024KB
  • RAID 5: Physical Disks 0:0:2, 0:0:3, 1:0:4, and 1:0:5
    • RAID Stripe Size: 8KB, 64KB, 128KB
    • Allocation unit size (a.k.a. block size): 4KB, 8KB, 64KB
    • Partition offset: 32KB, 64KB
  • RAID 10: Physical Disks 0:0:2 and 0:0:3 (Span 0); 1:0:4 and 1:0:5 (Span 1)
    • RAID Stripe Size: 8KB, 64KB, 128KB
    • Allocation unit size (a.k.a. block size): 4KB, 8KB, 64KB
    • Partition offset: 32KB, 64KB
• SQLIO v1.5.SG
  • File size: 64 GB
  • Threads:
    • RAID 1: 4
    • RAID 10 & 5: 8
  • Thread depth: 8
  • Buffering: Hardware
  • Test duration: 120 seconds
  • Wait between tests: 60 seconds

Test Harness 2: Dell PowerEdge 2950 w\ PowerVault MD1000

• Dell PowerEdge 2950
• Windows Server 2003 R2 Standard Edition Service Pack 2 (x64) patched with all updates available at test time
• 2 x Quad Core Xeon E5420 @ 2.50GHz (4 cores per CPU = 8 cores total)
• 4 GB RAM
• PERC 5/E controller connected to a PowerVault MD1000 drive enclosure operating in split bus mode
• Disk cache policies set to write back and no read ahead
• 146 GB 15,000 RPM Seagate Cheetah SAS hard drives
• RAID configurations:
  • RAID 1: Physical Disks 0:0:7 and 1:0:0
  • RAID 5: Physical Disks 0:0:7, 0:0:8, 0:0:9, 0:0:10, 0:0:11, 0:0:12, 1:0:0, 1:0:1, 1:0:2, 1:0:3, 1:0:4, and 1:0:5
  • RAID 10: Physical Disks 0:0:7 and 1:0:0 (Span 0); 0:0:8 and 1:0:1 (Span 1); 0:0:9 and 1:0:2 (Span 2) ; 0:0:10 and 1:0:3 (Span 3); 0:0:11 and 1:0:4 (Span 4); 0:0:12 and 1:0:5 (Span 5)
• RAID stripe size: 64KB
• Allocation unit size (a.k.a. block size): 64KB
• Partition offset: 32KB, 64KB, 1024KB
• SQLIO v1.5.SG
  • File size:
    • RAID 1: 128 GB
    • RAID 10: 512 GB
    • RAID 5: 512 GB
  • Threads: 4, 8, 16
  • Thread depth: 8
  • Buffering: Hardware
  • Test duration: 120 seconds
  • Wait between tests: 60 seconds

Test Harness 3: Dell PowerEdge 2850 w\ PowerVault 220s

• Dell PowerEdge 2850
• Windows Server 2003 R2 Standard Edition Service Pack 2 patched with all updates available at test time
• 2 x Dual Core Xeon @ 3.20GHz (2 cores per CPU = 4 cores total)
• 4 GB RAM
• PERC 4/DC controller connected to a PowerVault 220s drive enclosure operating in split bus mode
• Disk cache policies set to write back and no read ahead
• 146 GB 15,000 RPM Seagate Cheetah Ultra320 SCSI hard drives
• RAID configurations:
  • RAID 5: Physical Disks 0:0, 0:1, 0:2, 0:3, 0:4, 0:5, 1:9, 1:10, 1:11, 1:12, 1:13, and 1:14
  • RAID 10: Physical Disks 0:0 and 1:9 (Span 0); 0:1 and 1:10 (Span 1); 0:2 and 1:11 (Span 2) ; 0:3 and 1:12 (Span 3); 0:4 and 1:13 (Span 4); 0:5 and 1:14 (Span 5)
• RAID stripe size: 64KB
• Allocation unit size (a.k.a. block size): 64KB
• Partition offset: 32KB, 64KB, 1024KB
• SQLIO v1.5.SG
  • File size:
    • RAID 10: 512 GB
    • RAID 5: 512 GB
  • Threads: 4, 8, 16
  • Thread depth: 8
  • Buffering: Hardware
  • Test duration: 120 seconds
  • Wait between tests: 60 seconds 

Results
Here are the results of every test that I ran so that you can analyze all the nitty-gritty numbers for yourself. (Note that these are formatted for Excel 2007. If you don’t have Excel 2007 you can still view them by downloading the Excel Viewer from here)

• Disk Performance Hands On - PowerEdge 2950.xlsx
• Disk Performance Hands On - PowerVault 220S.xlsx
• Disk Performance Hands On - PowerVault MD1000.xlsx

Stay tuned…In Part 2 I will take a look at how partition offset, RAID stripe size, and allocation unit size affect RAID 10 performance.

Share This

• Twitter
• Google+
• Linked In
• Facebook

About Kendal

Kendal is a database strategist, community advocate, public speaker, and blogger. A practiced IT professional with over 15 years of SQL Server experience, Kendal excels at disaster recovery, high availability planning/implementation, & debugging/troubleshooting mission critical SQL Server environments. Kendal is a Senior Consultant on the Microsoft Premier Developer Support team and President of MagicPASS, the Orlando, FL based chapter of PASS. Before joining Microsoft, Kendal was a SQL Server/Data Platform MVP from 2011-2016.

Related Posts

Best Practices Performance SQL Server 2005 SQL Server 2008