Cloud, Data Center, News

Pure//Accelerate™ 2018 Conference News

I’m at Pure//Accelerate™ in San Francisco this week, and it’s shaping up to be an exciting event! The show features keynotes by Pure Storage CEO Charlie Giancarlo and guest speakers Jez Humble (who you may be familiar with from books such as The DevOps Handbook) and Dr. Kate Darling, leading expert in robot ethics and researcher at the Massachusetts Institute of Technology (MIT) Media Lab.

I’m particularly excited about this event because I’m a nerd, and storage is about as nerdy as it gets. But the conversations going on here at Accelerate are much broader than storage alone. The sessions I registered for this week focus pretty heavily on the cloud journey and on big data. I think the story Pure is telling this week is consistent with all the other storage industry mainstays: data is the new lifeblood of businesses, and without solid data storage technology, you’ll never be able to realize the full potential of the data you have access to.

The announcements coming out of Pure//Accelerate™ also show Pure angling to move up the stack and into a more strategic role within their customers’ organizations – this is also consistent with other major infrastructure players today. An infrastructure business is leaving opportunity on the table without some more direct hooks into the value of an organization; in Pure’s case, that means helping businesses derive value from the data that lives on their storage platforms. That brings us to the main announcement…

Pure’s Data-Centric Architecture

The world of data is changing quickly, and there are a number of changes that call for a different way of dealing with data. Things such as:

  • Capacity (we’re storing more now than ever before)
  • The core/edge computing paradigm, where we may need to bring some amount of storage to the edge
  • IoT and machine data is generated at serious volume and frequently needs to be processed in real time

“Our customers aggressively seek to use data to improve their customer experience and outdistance their competition,” said Charles Giancarlo, CEO, Pure Storage. “The unprecedented explosion of data coupled with new data center technologies enables Pure customers to build a data-centric architecture that enhances and simplifies their ability to use data for intelligence and advantage.”

Pure’s response to dealing with these new data challenges is summed up in their Data-Centric Architecture model, which has the following five core pillars:

  1. It is consolidated and simplified
  2. It can store and process data in real time
  3. It delivers storage on-demand and is self-driving
  4. It supports a multi-cloud model
  5. It is “ready for tomorrow,” which means it’s easily changed and adapted as technology and your requirements evolve

It takes a few big announcements to provide storage that addresses these pillars. That’s what Pure has given us this week at Accelerate.

Beyond the All-Flash Array

The two main models for delivering storage for all of recent history both have advantages and disadvantages. Networked storage (SAN/NAS) was pivotal in allowing the virtualization wave to wash over the industry, because multiple hosts could consume the same storage. It also provided opportunities for rich data services to be applied to the global data set within an organization, which increased efficiency and reduced cost. And finally, shared storage increased utilization by eliminating islands of independent storage. That said, it certainly caused some performance concerns (you may be familiar with terms like “noisy neighbor” and “IO blender” to describe this challenge). It also required significantly more expertise to deploy, manage, and troubleshoot.

The other model, direct-attached storage, has seen a resurgence, thanks mostly to big data applications. DAS is fast because of its locality; it’s also simple and cheap compared to networked storage, and suits analytics and cloud-y workloads quite nicely. However, DAS can’t offer the same reliability and efficiency benefits of shared storage systems, and managing DAS across hundreds or thousands of nodes is a real burden. So both networked and direct-attached storage have merit but drawbacks. Ideally, we’d like to have the best of both worlds.

In my opinion, storage of the future will have a form factor that looks a lot like networked storage, but will perform for applications more as if it were direct-attached. Thanks to NVMe-over-Fabrics becoming a reality, storage systems can be built today that take advantage of all the good reasons to share storage, but without the latency penalties that come from the network hop and the management overhead of DAS at scale.

The goal of NVMe-over-Fabrics is to provide access with no more than 10 additional microseconds of latency over the fabric, compared to locally accessing the NVMe. Additionally, the NVMe protocol requires less of the CPU than traditional SAN and NAS protocols, thanks to RDMA and a much leaner command set; this means that more CPU capacity can be dedicated to applications (compared to processing IO requests).

To fully take advantage of NVMe-oF, Pure is announcing the new FlashArray//X family of arrays which will ultimately replace the //M series of arrays. The //X family is all-NVMe and has the same potential to be disruptive in the NVMe-based storage market as the //M did in the all-flash array market. The line features five configurations to bring NVMe to all – from the NVMe-ready //X10 and //X20, to the all-NVMe //X50, //X70, and //X90, Pure’s densest, fastest array to-date.

The new all-NVMe //X family of arrays


On of the biggest draws to cloud computing is the consumption-based pricing model, and there are plenty of data center operators who would love to bring that financial efficiency to their data center. Even though they still had to buy in chunks and buy for future consumption, one option organizations have had to operationalize some of their storage costs is to lease the storage systems. Unfortunately, the IASB has really thrown a wrench into that plan by requiring that, as of 2019, operational leases longer than 12 months must be recognized on a company’s balance sheet (they typically haven’t been; they’ve only been disclosed). This change will throw all of the most important financial ratios out of whack. Suffice it to say that leasing your next storage array is probably not a viable option, so you’re looking at a hefty capital expense instead.

To allow customers to address both the leasing challenge and the desire to purchase IT resources in a cloud-y, consumption-based sort of way, Pure has announced this week that they’re extending the Evergreen™ Storage program to create a new offering called the Evergreen Storage Service (ES2).

ES2 delivers actual consumption-based, pay-per-use storage from Pure in your own data center. Pure brings out the storage and installs it; they maintain and upgrade it over time, and you just pay for it based on how many gigabytes you’re using. One of the cool things about ES2 is that you’re billed on a GB used basis. That’s as compared to some more established cloud storage services that bill on a per GB provisioned basis. Depending on how much overhead your organization tends to configure for safety’s sake, this simple difference in billing could save you boatloads.

“The Evergreen Storage program isn’t just a great customer service model, it’s a feat of engineering,” said Matt Harris, Head of IT for the Mercedes AMG Petronas Motorsport. “Pure has eliminated the most painful part of storage purchase and maintenance with arrays that were architected for the era of flash, cloud and machine learning. While legacy systems require full rip-and-replace upgrades every three-to-five-years, Pure has designed its arrays to easily integrate new features, capabilities and technologies – like NVMe – so that the benefits are passed along to our drivers immediately instead of several years down the line. In today’s highly-digital, fast-paced environment, Pure allows us to add competitive differentiation in real-time.”

Learn More

There are other exciting announcements being made this week too, such as additions to Pure’s AI-Ready Infrastructure in the form of the all-new AIRI™ Mini. To get more details from the show this week, consider visiting the home page to watch a live stream of the keynotes when they’re in session and pop over to the News section of the website for more details on the announcements this week.