NetApp goes all in with server side caching … again

In 2017 NetApp announced the acquisition of Plexistor during their annual customer conference in Las Vegas. Plexistor comes from the newly defined world of “Software Defined Memory” (SDM). They have recently rebranded the solution as MAX Data. MAX Data release is dependent on Optane DIMM supportability from Intel, which is expected in sometime late 2018.

NetApp has previously taken the position that as a way to continue improving performance they could take the data off Data ONTAP. An example of one of the ventures can be witnessed in the NetApp software FlashAccel. FlashAccel used ties in to VMware that maintained data coherency by requiring the data to first be acknowledged by the SAN prior to offloading to a PCIe flash device that resides on a host. In theory, the software would allow customers to offload reads from the storage array, providing additional headroom, and improve read latency for the application by keeping the reads local to the host.  Should the card fail the tiering would discontinue and reads would be directed back to the, potentially, slow disk. In reality, however, there was a tremendous cost associated due to the expensive flash PCIe cards required in every host in the cluster.  Even today pricing of the PCIe flash cards are greater than $1,000 per TB.

MAX Data, on the other hand, uses a proprietary Linux file system (ZUFS) to tier data from a block device to persistent memory. The persistent memory, in this case, will be from Intel’s 3D Xpoint Optane DIMMs on a physical host, connected via NVMe over Fibre Channel. It’s unclear how the tiering process or performance will be, as it seems the functionality is still a ways out. From a review of available information, we suspect that the Optane DIMMs will be able to facilitate read and write cache functionality due to the persistent nature of the memory.  The writes cached on the server would be exposed to unavailability in the event of the failure of the server, if the data that is cached in memory has not been tiered to the NetApp storage array. Moreover, the cost, just as previously stated in the FlashAccel example, could add up quickly with the requirement of adding 3D Xpoint DIMMs to support the anticipated capacity required for caching.

In a recent public Podcast it was enlightening to hear NetApp proclaiming this solution will be part of their “smashing orange” campaign. There were similar statements made during the SolidFire acquisition and throughout the all-flash FAS and new NVMe array announcements. We are honored by NetApp’s acknowledgement of Pure’s impact, and how we have encouraged our competitors to change business practices, introduced simplicity, adopting a customer-centric oriented model, and going flash! It’s flattering to hear that Pure is inspiring more innovation in the storage industry, even if it’s primarily for the sake of competing with us!

This move is a unique way for NetApp to increase performance without having to update the existing product line. The disparate way of addressing issues by throwing more product and layers of complexity is reminiscent of vendors NetApp had previously spent a lot of energy selling against with its Swiss Army knife approach (Data ONTAP). NetApp has always been proud of being able to sell a single solution / product to solve most customer requirements.  The benefits of having a single array that can take the place of several other products have undoubtedly put NetApp as a leader of the consolidated storage market, forcing others to create similar functioning products. NetApp now going down the road of adding additional hardware, software, management, and complexity is a clear diversion of its previous path.

In order for NetApp customers to take advantage of this new functionality they will need to be on a new all-flash platform (A300, A700, A700s, and A800), and have qualified NVMe-of networking in place. NetApp customers that purchased an all-flash FAS, with 6 years of up-front Premium support, have the ability to trade in their controllers toward a purchase of new controllers. The promotion, however, does not include customers that have aging hybrid or all HDD arrays. Thus, the NetApp customers that have anything other than the new all-flash storage controllers, will have to forklift in to the new platform prior to taking advantage of MAX Data.

Pure’s Evergreen storage business model enables existing customers to take advantage of future innovations to hardware and software, without the expense of additional capital to forklift out of the investment they’ve already made. Pure solutions continue to embody our philosophy of delivering simplicity, efficiency, and a great customer ownership experience. Our current FlashArray //M, or previous generation FlashArray, customers will enjoy the benefits of being able to non-disruptively upgrade to an all-NVMe array without any loss of performance, and without any data migration. Customers looking at Pure for the first time will know that whatever hardware or array software comes out in the future they will be able to enjoy the benefits with their Evergreen™ subscription to innovation.