IOPS Are Fun But Latency Wins

Posted by VIOLIN SYSTEMS on Apr 22, 2017 2:09:12 PM
You can Be Instrumental as a performance champion to your organization by revealing how lower latency, rather than higher IOPS, wins races in all flash storage for enterprise data centers. Like a driver behind the wheel of a Formula 1 (F1) racecar, you want to win that first race. Once you do that, your goal extends to winning other races and your aspiration becomes how do you become champion of the series.

Winning the series signifies that you and your team (pit crew, ownership, etc.) are the best at competing in tough and complex environments, which in F1 means a variety of races on various streets and roads in varying conditions.

In the data center, IT professionals translate the traits of a F1 driver into meeting performance service level objectives every minute of every day for the entire year.

Recounting Pivotal Outcomes in Storage Performance

The Latency Matters series features five articles and covered a lot of ground on the importance of lower latency in all flash enterprise storage. Here's a quick recap you and your team can race through:

  • Blog 1: The performance of all flash storage solutions is counter-intuitive as lower latency rather than higher IOPS drives database and application performance.
  • Blog 2: The primary driver of business value from all flash storage is not about how many transactions can be completed at the same time (IOPS), it's the consistency with which each and every individual transaction can be rapidly completed (latency).
  • Blog 3: Low Latency wins championships in storage environments with mixed and multiple workloads.
  • Blog 4: Performance is the most significant criteria when evaluating and selecting all flash storage; and IOPS, at a specific latency, is a key metric.
  • Blog 5: Lower latency translates to higher productivity, revenue gains, and compelling business value.

"Most enterprises have at least several applications where extremely low storage latency can directly translate to increased revenues, better customer experience, or differentiated competitive advantage."

"These applications tend to be transactional, database-driven or real-time analytics workloads where a 500 microsecond latency difference can be directly monetized."

"For these environments, AFA architecture can matter a great deal in an ability to consistently deliver sub 300 microsecond performance that companies can base their reputation on, regardless of widely varying workloads and operational workflows."

- Eric Burgener, Research Director in Enterprise Storage, IDC

Three Key Takeaways on Latency in Enterprise Storage

  1. Increasing IOPS theoretically allows more databases and applications to run on all flash storage, but all can become equally unresponsive due to higher latency.
  2. Doubling database and application performance with all flash storage involves reducing latency by half rather than doubling IOPS.
  3. It’s lower latency that matters most with all flash storage, especially in environments with mixed and multiple workloads.

Meet the Champion

We invited readers to join the race and enter the Need for Speed Sweepstakes throughout our season of Latency Matters articles. When the season concludes for the year, I will present our winner and crown a champion:

Position 1 - Champion | Driver: Jon L. | Team: Auburn Hills, Michigan

The Storage Grand Prix Continues

Ten years ago, Violin Systems revolutionized the storage market by pioneering all flash storage for the enterprise data center. On September 14, 2016, Violin once again reset the performance and economics of all flash storage by continuing our industry-leading innovation to enable the next-generation data center.

Click here to view the announcement webcast on-demand.