Blogs

Why Storage Performance Testing is Critical

By Jim Bahn, Senior Director, Product Marketing

If storage vendors could custom-build products based on individual customer application profiles, there might not be a need to test.  But they do not, and so storage engineers and architects must go beyond the simple performance profiles that appear in spec sheets, and the guidelines provided by “industry standard” benchmarks.  Customers must have data that’s relevant to their environment.  Testing is required because:

  • Storage workloads are different across different applications and platforms. No one would expect a streaming video application to stress a storage array in the same way as an email server.
  • Applications have wildly different requirements for storage performance. Latency for some applications has to be measured in microseconds, while for some, multiple seconds.
  • No one storage array can universally provide the best price/performance for all application types. Different storage systems perform differently for different workloads: there is no one storage system that fits all.
  • It is important to know how a storage system performs in your environment, including expected growth. What happens to performance if your workload increases by x%?
  • Industry standard benchmarks are known to be “gamed” by vendors who rightly want to present their products in the best possible light. Often, configurations are tested and reported on, which bear no resemblance to real customer configurations.
  • Vendors, in an effort to best their competitors, routinely publish specs that are not real-world in nature. A huge IOPS number may be offered, but it may have been achieved by using block sizes that are not representative of most applications. Or they may claim “random access” numbers based on a small dataset or data read by cache.  And there never is one block size for the composite workloads running on modern arrays.  It’s a distribution of sizes for different LUNs, reads and writes, and it changes over time.  Running a test showing only an average block size isn’t helpful to the customer.

 

Check out some of the customer case studies that show real-world benefits of realistic testing.  Go to Resources and look for Case Studies for Load DynamiX.  If you simply don’t have the resources to test, let us do it for you!