Everything you need to know about flash storage performance

Due to the unique characteristics of flash, performance validation testing is immensely challenging and critically important; follow these best practices

1 2 Page 2
Page 2 of 2

In Figure 2 below, we see the results. In this figure, we’ve sorted on the IOPS column to find the configuration that results in the greatest IOPS (approximately 22,014). Sorting by latency would quickly show figures exceeding 6ms for 500 or more concurrent workers.

Load DynamiX Enterprise output

Figure 2: A selection of the output report of the Iterator function, showing the effect of changing parameters on AFA performance.

Workload modeling goes to a greater level of detail. Whereas the goal of performance profiling is to test under a wide range of workload conditions, the objective of workload modeling is to stress the storage system under a realistic simulation of the workloads it will actually be supporting in production. Workload modeling requires a prerequisite knowledge of the characterization of the workloads, usually based on the storage engineer’s knowledge of the application and data typically provided by storage monitoring utilities.

The workload models should allow users to characterize access patterns with as much detail as needed. For example, block size can be represented as a realistic distribution of values, not only a single value. Different workloads should also be combined into a single “composite workload” that stresses different areas of the storage system. Testing a storage system with a workload simulation that is sufficiently realistic allows storage engineers to develop a great deal of confidence in their decisions about product selection and configuration prior to deployment in production.

These two methodologies represent the core of performance validation. They can be used to support several typical storage testing approaches. These include limits finding, such as determining the workload conditions that drive performance below minimal thresholds; functional testing (investigating performance and system behavior under a simulated load of various storage functions, such as backup, replication, and more; error injection, or the investigation under simulated load of specific failure scenarios; and soak testing, observing the storage system under load sustained over significant time (two days, one week, and so on). In short, the performance profiling and workload modeling test methodologies can be used to answer virtually any question one might think to ask about an all-flash array.

Flash-based storage arrays offer some tremendous advantages over disk-based arrays, but their fundamental differences make the storage buying decision more complex than ever. By following testing best practices in the lab, including performance profiling and workload modeling, AFA product selection and configuration become a mathematical exercise, not a guessing game.

Jim Bahn is senior director of products at Load DynamiX, a storage performance validation company. Jim ensures Load DynamiX products meet the needs of both storage vendors and IT organizations. He brings more than 30 years of marketing, engineering, and field experience from NetApp, Virtual Instruments, Hitachi Data Systems, and HP.

New Tech Forum provides a venue to explore and discuss emerging enterprise technology in unprecedented depth and breadth. The selection is subjective, based on our pick of the technologies we believe to be important and of greatest interest to InfoWorld readers. InfoWorld does not accept marketing collateral for publication and reserves the right to edit all contributed content. Send all inquiries to newtechforum@infoworld.com.

1 2 Page 2
Page 2 of 2