Everyone knows that solid state drives are fast. And while speed is great, consistent performance is essential. Why? Because datacenters are driven by service level agreements that demand the same level of fast, storage performance 24x7x365 – regardless of the workload complexity or changing IO intensity levels. So how can you evaluate consistent SSD performance?
SSD Performance Testing Best Practices
When evaluating enterprise SSD vendors and products:
- Ask vendors to have product performance claims validated by a proven, independent, 3rd party auditor
- Make sure that the test methodology used leverages enterprise workloads that include predominately random IOs/second and use both query and update operations (Online Transaction Processing (OLTP), email, and/or database applications deliver these types of workloads)
- Confirm that the test evaluates and reports on user responsiveness (eliminating excursions) and latency – with respect to changing IO intensity levels (This is an important performance consistency indicator as it shows the level of variance in end user response time)
- Ask the vendor to disclose the test duration (1 hour tests are good to gauge general performance, but insufficient to prove long-term consistency. A 4 hour test time is needed to gather a sufficient amount of raw data that can confidently be extrapolated to prove long term performance consistency)
- Ensure that the performance test is repeatable, reproducible and uses consistent test metrics (This enables easier comparisons across vendor products)
- Finally, request that test results be made publically available by the auditor
SPC-1C Performance Benchmark
The Storage Performance Council’s SPC-1C storage benchmark is one way to evaluate sustainable device level SSD performance and consistency. The SPC-1C workload is comprised of a set of I/O operations that are designed to demonstrate the performance of a small storage subsystem while performing the typical functions of a mission critical application. While enterprise workloads vary, the SPC-1C test uses synthesized scalable real world workloads that represent a segment of enterprise applications characterized by predominately random I/O operations and require both query and update operations (for example: OLTP systems, Database systems, and/or Mail Server applications).
The SPC-1C test methodology is repeatable and reproducible because it is clearly defined, platform independent, and supports popular operating systems such as Windows, Linux. This enables “apples to apples” comparisons across vendor products and establishes a level “playing field” for all SSD manufacturers. It also measures and provides trending data on user responsiveness and latency with respect to changing IO intensity levels. Finally, the Storage Performance Council is a proven, respected storage industry organization that also provides independent audit services and publicizes storage performance benchmark results online.
Go check it out here.