Evaluate Pure Storage

Performance runs deep at Pure Storage.

Our performance philosophy is simple:

100% flash.
0% regret.

Disk is great for backup and 2nd-tier applications, but for the applications that power your business, you need flash.

Learn More

latency is king.

For flash to be transformative, it must be predictably fast Pure optimizes for consistent latency, even through failure.

Learn More

Optimize for
the real world,
not benchmarks.

Although 4K benchmarks might produce nice IOPS numbers, that isn’t the real world. Pure optimizes for varied IO sizes and types, and mixes them with ease.

Learn More

Bet on
Moore’s Law.

Unlike competitors who attempt to optimize via ASICs and proprietary architecture, the Pure recipe is software + x86. We think history’s on our side.

Learn More

All-Flash vs Tiering and Caching.

Don't gamble with performance by mixing mechanical disk and flash.
Compare the performance examples to see why:

PureStorage vs Caching and Tiering

Disk arrays with flash caches have a traditional “bi-modal” IO distribution: Some of the IOs are serviced by the array’s DRAM and flash cache or tier (and thus delivered in 100s of microseconds), and some fall-through cache to mechanical disk (delivered in 10s of ms, two orders of magnitude more slowly than flash). What percentage of IOs hit flash is dependent on the cachability of the workload, the predictive ability of the controller software, and your hand tuning – and therein lies the problem: on an IO-for-IO basis, you never know if you will hit flash or disk. And it is these outlier disk IOs that your applications are gated by. Consider that the typical database transaction is made up of several storage IOs: if any of those IOs hit disk, the containing transaction performs like disk rather than like flash. The Pure Storage FlashArray allows you to “roll up” that long tail of outlier IOs, delivering predictable sub- millisecond latency that you and your applications can count on. (Pure Storage costs less than a legacy disk.)


Why Average Latency Lies

Watch now

Pure Storage 5-Minute Demo

Watch now

Don’t be fooled by the 4K IOP charade.

What really matters is how a storage array will perform on your application.

Average Pure Storage FlashArray IO size:

51.54 K

(averaged across all call-home data from Pure Storage Support)

Last Updated: 11:21:03PM PDT

Pure Storage implements a variable internal block size, meaning that IOs ranging from 512bytes to 128K are all processed as “one IO” inside Purity

Competitive 4K-aligned flash appliances break-up all incoming IO into 4K chunks, yielding a linear drop-off in performance with IO size. 1 64K IO = 16 4K IOs

The industry commonly markets a 4K IOPS benchmark to make numbers look high, but real-world environments are dominated by IO sizes of 32K or larger.

Pure Storage has optimized the FlashArray for the real world, not the 4K vanity benchmark.

We’re betting on Moore’s Law.

Software + x86 architecture helps us deliver predictable, repeatable performance and capacity growth, year after year.

We’re betting on Moore’s Law.

Some flash products have gone the route of specialty ASICs or FPGAs and unique/proprietary hardware innovation to try and deliver breakthrough performance. Pure Storage has a different philosophy: bet on the Intel curve. The Pure Storage FlashArray has an inherently CPU-bound architecture, meaning that we’re able to improve performance in-step with CPU improvements driven by Moore’s Law. We’ve had a history of doing so since our inception, and we don’t see any slow-down in sight!