By Eric Burgener – Research Director, Storage, IDC
External storage revenues have largely shifted in the last several years to flash optimized platforms, which include both Hybrid Flash Arrays (HFAs) and All Flash Arrays (AFAs). With evolving 3rd Platform computing workloads, which include a mix of legacy as well as next generation applications running on virtual infrastructure, it is clear that flash has to be part of the primary storage mix.
Relative to hard disk drives (HDDs), raw flash capacity is still much more expensive and vendors have differed on the best way to incorporate flash. Some vendors introduced enterprise-class arrays built entirely out of flash media, using data reduction and other technologies to bridge the $/GB cost difference, whereas others built systems with a mix of both flash and HDD, using flash optimized write and intelligent caching or tiering algorithms to maximize the amount of I/O serviced by a relatively small flash layer. Performance between particular HFA vendors can vary significantly, depending on the actual implementation and amount of flash optimization. Given the mix of applications that most enterprises need to serve, vendors from both sides can field credible arguments for the use of their systems as a general purpose storage array.
Changing market dynamics are affecting what customers are looking for in flash-optimized storage though. As flash raw capacity costs have come down, more customers are thinking about using an AFA as a general purpose primary storage platform, moving the deployment model for them away from the historical “dedicated application” approach to more of a “mixed workload consolidation” approach. As this occurs, new considerations arise around cost-effective consolidation of both primary and secondary storage applications onto a single platform and replicated configurations.
For the most performance sensitive applications, the use of AFAs has been easy to justify. As flash costs come down, it becomes financially viable to move less performance sensitive workloads to an AFA to take advantage of unified management and the secondary economic benefits of flash deployment at scale. But on the continuum of most to least performance sensitive applications, there are generally many more less performance-sensitive applications.
For those customers that want to place both primary and secondary applications on a single platform, an HFA that supports volume pinning, tiering, or some other form of persistent storage pooling provides a great opportunity to co-locate many different types of applications. These HFAs can deliver all-flash performance all the time for applications that need that, and more cost-effectively provide capacity for those applications that need that – all on the same platform. This is one of the reasons that IDC expects the HFA market – although it is growing slower than to the AFA market – to still be three times the size of the AFA market in 2019.
If customers are hosting mission-critical consolidated workloads on a general purpose array, they often want to be able to replicate to a DR site. Even if those customers want an AFA at the primary site for performance reasons, they often want a less expensive HFA system as the replication target. If that target system supports persistent flash storage, this gives those customers the opportunity to still deliver all-flash performance for a select number of the most critical applications in a post-failure scenario, but it also means they enjoy a lower cost system during the 99.9% of the time that they will not be using that system as a primary. Many AFA vendors cannot, however, replicate to any other system than their own. In the recent IDC AFA MarketScape, published in December 2015, vendors who supported a way to replicate to heterogeneous storage platforms – even if it was an extra cost item – received credit that vendors who could not do that did not get.
What this all boils down to is that customers that buy AFAs as a general purpose primary storage platform have very good reasons for doing that, often choosing to maintain a second, lower cost platform for secondary storage. And customers that buy HFAs (which are generally always being used as a general purpose storage platform) have very good reasons for doing that. This clearly points up the most “customer friendly” portfolio approach that enterprise storage vendors can pursue: offer both and let the customer decide which platform they want to deploy. Vendors that only offer one or other platform are put in the position of explaining why that one platform is best for everything, while those that offer both can just default to the disarmingly simple “you choose” approach. In the aforementioned IDC AFA MarketScape, those vendors that could offer both were rated higher than those that could only offer an HFA or an AFA.
There is one final point to take into account. When considering a portfolio approach to crafting complementary primary and secondary storage solutions, unified management can be another very important consideration. Even if two different systems (e.g. an AFA and an HFA) offer different performance and cost characteristics, if they can both be managed using a common set of monitoring tools, data services, and other administrative operations, this provides additional advantages that are very important to certain types of buyers. As good as offering both an AFA and an HFA is as part of a portfolio strategy, that approach is even better when both systems leverage the same operating environment (with perhaps a few optimizations built in that are just for the all-flash environments) and can be managed with the same familiar management.