Flash is taking over the world. With prices falling, we are now at a tipping point where flash is the leader rather than the upsell in new storage acquisitions for the data center. However, opinions differ regarding where this flash should be inserted. There are three main sides to this argument, and at the base of it all, it comes down to where you believe your performance point should be.
The all-flash array, a relative newcomer to the world of storage devices, follows the tried and trusted format of a shared array in much the same way that traditional storage has been delivered and consumed for the last twenty years or so. Ardents like Pure and Nimble, and those of the old guard like EMC, NetApp, and HP all believe that storage performance should be in the array, where the data is stored.
As for the hyperconverged camp, it is a little like a beatnik, a trendy hipster. Its proponents are the cool kids on the block. The hyperconverged kids believe that the days of the monolith storage array is over. They are committed to a brave new world of bringing storage back locally to the compute layer, but with a twist. They are sharing their local disks and SSDs among themselves in a commune hippie–type way to mitigate the shortcomings of local storage.
The final side in this triangle is server-side acceleration. These are the mavericks or the enlightened, depending on your viewpoint. Server-side compute advocates believe that performance should be in the compute layer, but they also believe that storage should be shared. They are arguably more “love and peace” than the hyperconverged crew, as they believe that the monolithic arrays still have a place in the world—not to deliver performance, but to deliver on their promise of shared storage. They even have love for the hyperconverged camp.
In his paper titled “[tvp_pdf name=”TVP: Storage is a royal pain in the derriere”],” downloadable from The Virtualization Practice, Tom Howarth investigates these three paradigms, highlighting the relative strengths, weaknesses, risks, and constraints of each technology and examining where, if anywhere, they will be in the next couple of years.