Data Center Real World Storage Workloads

Library Content Type:
Publish Date: 
Monday, January 1, 2018
Focus Areas:

New IO Capture and analysis tools - such as those available for free at (TMW) - now enable users to capture, analyze and test Datacenter Real World Storage Workloads
(RWSWs). RWSWs are very different from synthetic lab workloads. RWSWs are comprised of constantly changing combinations of IO Streams and Demand Intensity whereas lab workloads apply a fixed and constant workload. Datacenter storage performance depends on how well storage responds to these dynamic RWSWs and often does not align with published manufacturer specifications or single access pattern corner case tests. IO Capture of RWSWs allows users to understand what IO Streams occur at the Block IO level and to analyze in-situ server storage performance that occurs during the IO Capture process. RWSWs also allow the user to validate software stack optimizations and to qualify server storage to the actual IO Streams that occur. This paper uses the TMW IOPofiler toolset to capture and analyze a Block IO level capture on a 2,000- outlet retail chain web portal. The 24-hour IO Capture file is processed to extract SQL Server application IOs which are then used to present in-situ SQL Server performance. The extracted SQL Server workload is also used as the test stimulus in RWSW tests that compare storage performance of a sample pool of Datacenter class storage.