Are You Using the Wrong Metric to Define Your Next Storage Array Purchase?

laptop connected to servers

Posted on Tuesday, June 5, 2018

When it comes to comparing storage hardware from different vendors, the industry has settled on Input/Output per second– IOPS– as the metric of choice. On one level that makes sense: the speed at which data is written to disk will have a significant bearing on the overall performance of your applications. 

But there’s a problem with IOPS as a metric: the results may not be as meaningful as you hope.

IOPS are hugely variable

If you’ve ever owned a notebook computer or smartphone, you’ll be well aware of the battery life statistics. The manufacturer’s claims rarely match your own experience because their tests are performed under optimal conditions, allowing them to achieve exceptional results.

The same is also true of data storage arrays and IOPS. Manufacturers perform all of their IOPS testing under optimal conditions to ensure that they can publish market leading statistics.

So just like your cell phone battery, array performance may be very different in the real world.

Design your own metrics

If IOPS can’t be trusted, what metric should you be using to assess performance? Just as your IT infrastructure is specific to your business, so too are your performance metrics, which means that you will have to design your own. 

Ultimately, the only metrics that matter, however, are performance of your workloads. Raw IOPS values will not adequately define whether an array is performing or not.

Instead, try assessing performance in terms of your own real-world workloads. If the system is hosting SQL databases, for instance, try measuring the maximum number of concurrent SQL users. If the application is VDI, test the maximum number of instances you can run before performance degrades beyond acceptable limits. 

Whatever metrics you choose to benchmark performance, make sure that they adequately reflect your workloads. And when you next upgrade your arrays, ensure that the vendor is able to demonstrate performance according to your workload needs, not just their headline IOPS rate.

For help with maintaining and optimizing your existing arrays, contact us today.

Download article as a PDF - Are You Using the Wrong Metric to Define Your Next Storage Array Purchase?

More Articles

Server cabinets in a row

Delaying a Legacy System Upgrade? You’re Not Alone

Retaining legacy IT systems is deeply unfashionable, but many businesses are doing it anyway.

Server against rainbow background

Magnetic Storage Is Even Older Than You Thought

A new scientific discovery shows that magnetic storage goes back long before tape.