One of the headlines at this years Flash Memory Summit was the much broader participation of the traditional storage vendors. Dell, IBM, Hitachi and HP were all there. Not to be left out, EMC also had a presence and we got a chance to sit down with them and catch up on how their XtremIO strategy is progressing.
As a recap, EMC bought XtremIO about three years ago. We predicted at the time that this was going to be a good move for EMC despite the fact that XtremIO needed some more incubation time. But EMC stayed focused and remained patient, delivering the product to the general market last November.
What is XtremIO?
Architecturally XtremIO is a scale-out design, which means that it expands in a nodal fashion. As each node is added, performance and capacity expand with it. The product’s primary claim is to deliver very consistent performance and consistently low latency, while providing a set of data services that function at all times. These services include thin provisioning, deduplication, snapshots, as well as their own organically developed RAID variant that is custom designed for flash storage. XtremIO recently announced the addition of compression, which is something we think is a key to use in conjunction with deduplication.
XtremIO is ready to address what we at StorageSwiss are seeing as an emerging buying pattern for all-flash arrays. The first phase of an all-flash purchase is to address a specific problem in the environment. Typically an application owner or maybe even customers are complaining (loudly) about performance. For this they often select a point solution. They just want the pain to go away, fast.
In the second phase, storage planners realize there is additional performance headroom on the all-flash array and will then put workloads with a similar I/O pattern on the device. The third phase typically consists of consolidating a variety of workloads on the all-flash array regardless of the I/O pattern.
The problem is that each of these phases often requires a different storage system. Phase one is ideal for a simple scale up system. Phase two can be a scale up system with expansion or can start to justify a scale-out system. Phase three starts to really require a scale-out system that can provide consistent high performance and low latency.
XtremIO has often been thought of as a scale-out solution but it can also be delivered in a single node configuration. As a result, it can meet the needs of phase one. In fact, EMC confirmed that over 50% of their initial purchases are a single node. They also recently announced a smaller starter node that is filled with half of the potential storage capacity. This lowers the initial entry price to solve the phase one performance problem while allowing the customer to meet the growth required in phases two and three.
As phase two matures, the customer tends to hit either a capacity, quality of service or performance wall. Consequently, phase three requires that performance and capacity grow to meet all of the future demands of the environment. XtremIO by its scale-out nature, should be able to address the initial phase two concern and its consistency of performance and low latency, allows it to also meet the long term needs inherent in phase three.
EMC could have been successful in the flash market simply due to their sheer presence and large market share. But XtremIO is illustrative of a more strategic approach to being a player in the space. EMC claims they have 500 revenue customers using XtremIO for production workloads since November 2013. That is a rapid initial launch. While there is still more software improvement and product maturity to be had, XtremIO’s first 8 months show they seem to be on the right track.