Recently, Nutanix got awarded a key patent for software-defined storage, and it got me thinking, “What is SDS?” Well, to me, one key aspect of SDS is the fact that it is hardware agnostic. Therefore, by default, Nutanix is not a provider of SDS. Now, don’t get me wrong: Nutanix is an awesome product and quite innovative in its offerings.
But, software-defined storage it is not. Hyperconverged? Yes.
So, what are the differences between these two terms?
Software-defined storage is a platform-agnostic presentation of storage. Examples include Joyent, Nexenta, and SolidFire. One of the key tenets of their offerings is that you do not have to purchase hardware to run your storage software from them. There may be reference architectures and recommended practices on how to build a more performant array using their software, but you are not compelled to purchase a particular type of hardware from them. This is where traditional SAN vendors (EMC2, Hitachi Data Systems, NetApp, etc.) differ: you must purchase their hardware to run their storage operating system and application. A major advantage of a true SDS product is that of upgradeability. There is no nasty lifting and shifting of your legacy hardware with the resultant risks that will bring. You need more performance? Just swap out your control heads one at a time with new, faster, standard x86 servers running your SDS software of choice. There is no glass ceiling. Another benefit is a serious opportunity for real cost reductions in your storage budget, due to the ability to use utility whitebox hardware servers and JBODs (“just a bunch of disks”).
Hyperconverged products have storage, compute, and network all in a single box, with a built-in hypervisor. Examples include SimpliVity, Pivot3, and Nutanix. All of these vendors sell their secret sauce in their own proprietary hardware. You cannot purchase their software out of the box and bake your own version; therefore, you are tied to their definition of what is best. A benefit over traditional storage is that they tend to be scale-out rather than traditional storage vendors, which, when you reach your vendor-defined glass ceiling, make you lift and shift for a larger model, discarding the previous model.
Now, you may ask, “Why aren’t Vblock and FlexPod in this section?” Well, to me they are not hyperconverged; they are nothing more than a set of reference architectures that you can purchase as a single product number. Further, they too have glass ceilings with regard to growth capacity and performance.
OK, so what about vSAN? vSAN is VMware’s Virtual SAN, which hooks into the vKernel of a set of three or more ESXi hosts and utilises local storage, SSDs, and traditional disks to present a pool of storage to the ESXi hosts. Now, the vSAN is an SDS product, as there is no predefined hardware that you must purchase. However, it is also hyperconverged, when, as it must be, it is coupled with ESXi on a host.
Perhaps it is time for a new paradigm: SDHC (software-defined hyperconvergence).