Curious Case of Disaggregated HCI or dHCI
The world of tech is an interesting one.
As a start up, you get the weird stares and nobody cares too much for you. Flourish and do well, competition takes notice and starts to throw FUDS (fear, uncertainties and doubts). Some try to find new use cases for their deprecating solutions. When you go mainstream, realisation that it's probably too late to re-engineer and join the party sends everyone into a frenzy to acquire. And for those that missed the 'fire sale', they work hard at trying to score an invite to the party with clever marketing, repackaging existing solutions and calling it the same.
I'm eluding to the inclusion of a new category recently called dHCI, also known as disaggregated hyper-converged infrastructure. Official definition states that dHCI is a variant of HCI that allows independent scaling of storage and/or compute, at the same time staying true to the values of the original HCI.
Look closely and start peeling back the layers, we start to notice the commonalities among these so called 'dHCI' solutions, and something familiar emerges from the shadows that look a lot like the traditional SAN.
Coincidentally the “storage components” of all dHCI's started life as a traditional SAN. It was never built natively and specifically for virtualised workloads from the go. In order to deliver 'hyperconvergence', it often leverages existing featuree within VMware or the hypervisor. For example, the use of VVOL for storage provisioning and management. A quick recap, VVOL (which have been around for some years with limited deployments) allows storage to be provisioned and managed thru policies, much like vSAN. However, its capabilities were heavily dependent and limited on individual storage hardware/vendor, not to mentioned they were proprietary. Other dHCI providers created plugins to manage their storage stacks from within vCenter, again something that existed since the beginning of time.
Remember the first generation of Converged Infrastructure or CI that predates HCI?
It consists of a combined bundle of SAN, SAN switches and servers, housed in a neatly cabled rack (hint: VCE). All best-of-breed in their own domain, and loosely integrated with a nice GUI setup. I guess its suffice to say, the fact that VCE no longer exists seem to suggest that it didn’t go too well. Still many vendors kept trying for years (improving bits and pieces of that fancy GUI) holding strong to the belief, that this was what customers needed.
Don’t get me wrong, I'm sure there’s some bells and whistles that make dHCI differentiate itself from native HCI, but I can’t help but think it's a repeat and forceful way to repackaged a bunch of legacy kit and call it new. The VVOL integrations and vCenter plugins can be had today with traditional SAN, tried and tested for years. Why would anyone then consider a stack that is 'new' that looks a whole lot like 'old'?
If I'd be fair, the one area of innovation (if u can even call it an innovation) I can think of is that some dHCI solutions tries to redesign/rearchitect the hardware chassis to fit both storage nodes and compute nodes. And even with that, there is a rigid configuration as to how many storage nodes / compute nodes is required to start a viable system.
HCI have always been about championing the notion of the Software Defined Storage. The key principals were built around the power of choice, commoditisation of hardware and the vision of a software defined data center. Some would also agree that HCI forms the foundation and are building blocks towards their journey into the hybrid cloud.
So the question remains for those evaluating dHCI solutions. Do you consider dHCI revolutionary or just another case of history repeating itself? Or perhaps a case of a vendor grasping at straws of a lost opportunity…
“Our lives are defined by opportunities, even the ones we missed…”
(from the movie Curious Case of Benjamin Button)
