Introducing Petabyte-Scale Flash Storage for Cloud IT

A little over a year ago, we introduced our flagship all-flash storage array – FlashArray//m. We’ve continued to innovate and re-imagine every aspect of the storage infrastructure to deliver storage that is effortless, efficient, and evergreen – three key pillars that are foundational elements for delivering cloud-like simplicity and agility, yet with enterprise-grade resiliency to implement the new cloud operating model of IT. We call that Smart Storage for Cloud IT.

Screen Shot 2016-10-11 at 12.56.41 PM

You can learn more about Smart Storage here

When we launched the //m, everyone asked us what the “m” meant.  We had some fun with that – it was mini for it’s amazing 3U size, it was mighty for it’s capacity and performance…but perhaps most importantly we said it was modular – the //m was designed for taking advantage of rapid controller and flash upgrades, enabling customers to grow in an Evergreen™ fashion without downtime.  Here’s the slide we used to explain that when we launched:

Screen Shot 2016-10-11 at 1.02.29 PM

We also made an important promise – we said we were committing to a rapid pace of upgrades (both hardware and software), so customers could ride the flash innovation curve with Pure.  Today we make good on that promise, with a whole host of new upgrades for //m which focus on higher scale, higher performance, higher density, and higher resiliency.

Bigger, Faster, and Denser //m:

  • Petabyte-scale FlashArray//m – now FlashArray systems can scale to 1.5PB effective, in just 7U!
  • 3.8TB and 7.6TB Flash Modules – or 11TB and 23TB effective flash modules after typical data reduction, if you prefer 😉
  • 30% performance increase across the family – new Intel Broadwell controllers bring a 20-30% performance increase to //m20, //m50, and //m70
  • Software scale for existing //m’s – our software-defined architecture enables us to improve existing arrays without upgrading hardware too.  All the first-generation //m arrays also get increased capacity scale by up to 50% via upgrades to our Purity Operating Environment (OE) version 4.8

Higher Resiliency:

  • FlashArray//m has achieved 99.9999% availability – we’re announcing that we’ve hit the 6x9s mark for FlashArray//m in our first year of shipping, and we’re raising the bar on how that’s tracked and calculated
  • Always-on QoS – new zero-knobs, always-on noisy-neighbor protection augments our high and consistent all-flash performance for safe consolidation
  • NPIV – support for host-transparent upgrades so server and VM teams never know upgrades are happening, makes FlashArray upgrades even more painless

We also launched a fun new web series called “Outside the Box” that is designed “for techies, by techies” to share tech deep dives and innovation stories. We’ve got innovation constantly pouring out the doors of Pure – so we wanted a more frequent way to share with everyone what we’ve got cookin’ in the lab.  The inaugural episode features today’s product announcements and more. Click here  to get in on the fun and catch up on today’s announcements.

Next-Generation, Petabyte-Scale FlashArray//m

The newly updated FlashArray//m, the fifth generation of our flagship FlashArray, now scales up to 512 terabytes of raw flash, which translates to approximately 1.5 petabytes of effective capacity – all in just 7U of rack space.

Screen Shot 2016-10-11 at 5.35.14 PM

Leveraging Pure’s Evergreen architecture, the FlashArray//m  allows customers to take advantage of rapid technology advances in both compute and flash. The new //m70 is equipped with 80 Broadwell CPU cores per system, 3.8TB and 7.6TB 3D-NAND Flash Modules, and 2TB DRAM for delivering zero-compromise data services with scalable metadata.

Why does this matter?  In a word: consolidation.

Screen Shot 2016-10-11 at 1.17.07 PM

In today’s cloud IT architecture consolidation matters, and the Petabyte-scale //m was designed to safely consolidate a wide range of workloads, enabling you to reduce your complex storage tiers and just put everything on flash.

It’s also all about taking advantage of flash media advancements without losing your FlashArray investment.  Customers are now starting to internalize how fast flash changes, and how important it is to pick an evergreen architecture for an all-flash array.  Just look at how quickly even Pure’s product line has evolved:
Screen Shot 2016-10-11 at 5.37.40 PM

If you wanted to buy 1PB of flash from Pure when the FlashArray FA-300 first came out in 2012 it would have spanned 6 racks!  With today’s announcements, 1PB occupies just 5U, and you can guess that it will be even smaller next year 😉

The FlashArray//m delivers one architecture that scales from TBs to PBs. Customers can start at any size and scale online to petabytes, non-disruptively, with no performance impact, all without data migrations or re-buying hardware or software.

We’re also introducing new //m20s and //m50s. Across these new controller options, the FlashArray//m delivers a 20 to 30 percent performance boost and a 100 – 276 percent capacity boost over the first FlashArray//m generation.

In addition to the new petabyte-scale //m, we are increasing capacity scale on existing //m20s, //m50s, and //m70s by up to 50%. With Purity OE 4.8, we’ve also improved object scale by 10X, enhancing overall system scalability and enabling large-scale Copy Data Management with up to 50,000 space-efficient snapshots on the system. What’s really cool is we’re doing these increases without even changing the hardware – this is all via software enhancements. So customers that bought into our Evergreen model can once again get storage innovation without a refresh. This is Evergreen in action!

So what has this Evergreen experience looked like for customers that bought Pure early?  A customer who started with Pure in 2012 with a modest array could have scaled non-disruptively their capacity by 53X, their performance by 5X, and traversed through four generations of Intel processors – all without a wasted investment!

Five_years_of_Evergreen

 

At Pure, Evergreen isn’t just a marketing slogan – it’s a completely different technology platform and business model. We’ve been delivering it for years and we’re going to continue to do so – so you can gain a competitive edge in your business.

The new FlashArray//m models are shipping and available to order now, and Upgrade Flex bundles enable existing FA-400 Series and first-generation FlashArray//m customers to upgrade and receive trade-in credit for existing controllers if/when that becomes necessary.

Reliability: Still Job #1

As customers consolidate dozens or even hundreds of applications, they don’t just need bigger all-flash arrays – they also need arrays that are both efficient and consistent – and that takes software.  Flash has to be economical for large-scale consolidation, and customers need the peace-of-mind that consolidation is safe – both from an availability and performance perspective.

Screen Shot 2016-10-11 at 1.35.53 PM

Pure has long been a leader in efficiency with our best-of-breed data reduction technology, and our new //m arrays take that efficiency even higher.  But we’ve also worked very hard on reliability and performance consistency, and that’s the theme for our next few features.

Always-On QoS

As large-scale workload consolidation becomes the norm, customers worry that “noisy neighbors” – applications that starve the performance of other applications on an array – might emerge. Typical QoS implementations to prevent noisy neighbors are complex, requiring per-volume min, max, and burst IOPS to be set and then constantly adjusted as application needs evolve.  Could you imagine making these decisions for every storage volume in your environment (this is a Volume provisioning wizard from a competitor’s QoS implementation), as an example?

QoS_example

 

Ugh, complicated. What’s worse, improper QoS configuration can cause more harm than good, and can unnecessarily cap a workload’s performance even if the array has plenty of headroom!

The FlashArray//m’s new Always-On QoS feature requires zero configuration, so any customer can benefit without the complexity of legacy QoS implementations.  How do you implement Always-on QoS?  Easy – just create a volume:

Screen Shot 2016-10-12 at 9.25.08 AM

First – let’s state the obvious, it’s always-on, like all of Pure’s services (data reduction, encryption, snapshots, etc.). Any application can burst up to the maximum utilization of the FlashArray, as long as it is not impacting other applications. If the application becomes a “noisy neighbor” and begins to impact other applications, Always-On QoS will throttle the problem application down to ensure all other applications get their fair share of array performance.

 

Words are nice, but demos of a feature like this are much more useful – so here’s Pure’s performance professor, Lou Lydiksen giving a demo of the new Always-on QoS:

But what about large-scale service providers who want to set per-volume QoS so they can charge customers differently for tiers of service?  Architected with a sophisticated IO credit system under the covers, Always-On QoS is simple and autonomous, making it ideal for most use cases, but it builds the foundation for Pure introducing more fine-grained, policy-based QoS in future releases – stay tuned!  With this feature we’re democratizing QoS for the masses, but this is only the beginning.

Always-On QoS is now available in an Early Access Program for select customers with General Availability for production workloads planned for Q4 of calendar 2016.

NPIV Support for Host-Transparent Upgrades

Network Port ID Virtualization (NPIV) support on the FlashArray//m makes the non-disruptive software upgrade experience even more robust and invisible to the host and server and virtualization admin teams. With NPIV, IO intended for ports on a FlashArray//m controller that have been temporarily restarted during the software upgrade process is transparently directed to the other FlashArray//m controller, without any dependency on host multipathing software. NPIV removes the risk of downtime due to misconfigured host multipathing software – a common worry for large enterprise environments – and makes storage software upgrades transparent to server administrators by eliminating IO error alerts that would otherwise be generated at the application host layer.

Six Nines of Availability, Redefined.

In its first year of shipment, the FlashArray//m has achieved 99.9999 percent availability across the installed base, which equates to only 31.5 seconds of downtime on average per year, demonstrating the FlashArray//m can deliver the level of resiliency needed for mission-critical enterprise applications and always-on clouds.

Screen Shot 2016-10-11 at 1.55.45 PM

Pure Storage is also redefining the way availability is measured and reported to align directly with customer expectations and experience of availability. FlashArray//m’s measured 99.9999 availability is inclusive of upgrades and maintenance, across both hardware and software, and without requiring a second array configured with replication (read the fine print on some competitor’s data sheets!).

In addition to the resiliency architected into the FlashArray//m, 99.9999 percent availability is delivered with Pure1® Global Insight, our cloud-based predictive analytics.

Screen Shot 2016-10-11 at 2.04.20 PM

Since its introduction almost a year ago, Pure1 Global Insight has helped to automatically identify and resolve thousands of issues, and prevent over 170 Severity 1 issues to date. Pure1 Global Insight is a key ingredient for delivering an always-on experience to Pure Storage customers, as the install base continues to grow.

What’s Next?

Taken together, today’s announcements take a radical leap forward in delivering Smart Storage for Cloud IT. Storage that is effortless, efficient, and evergreen.   The new FlashArray//m coupled with Purity OE 4.8 mark substantial moves forward in scale, density, performance, resiliency, and consistent application performance.  Today’s announcements also highlight the 3-pronged innovation strategy we invest in here at Pure:

Screen Shot 2016-10-11 at 2.14.27 PM

Today we delivered a massive improvement in performance and scale – by coupling rapid hardware iteration with software enhancements.  We made most of that innovation backwards-compatible to existing customers with a simple software upgrade, and our Evergreen architecture enables everyone to upgrade their hardware to the latest and greatest non-disruptively if they want to stay on the sharpest edge of the curve.  We added a number of new software features to drive resilience, and we showed how we’ve been able to combine this with “IoT” big data intelligence to continue to drive-up reliability in our platforms.

But of course, we’re not done yet!  If you enjoyed today’s Outside the Box episode – stay tuned for the next episode for a deeper dive into a technology we believe will have a profound impact on all-flash arrays in the coming years: NVMe.