Recently, it was brought to my attention that some do not consider Cohesity to be “hyperconverged.”
. @datacenterdude really? i think industry wide the term hyperconverged has a specific meaning. This is just 'convergence'.
— Duncan Epping (@DuncanYB) May 12, 2016
After some back and forth, I decided to state my positions in detail. I’m going to stand by my position that it IS hyperconverged, but first we need to take a look at both CI and HCI and layout the differences in architecture.
VBlock, FlexPod, FlashStack. To me, these are examples of Converged Infrastructure. Disparate infrastructure components cobbled/cabled together in a go-to-market package, sold as a single SKU, and sometimes having a layer of management software on top to orchestrate the entire stack.
As someone who was intimately familiar with FlexPod over the last few years, the pain point that led to the solution was mostly around purchasing power in a modular fashion with pre-configured installations (roll it in, plug it in, begin using it). There is also an angle around “one phone number” support across multiple stack vendors, essentially a “Batphone” for your CI stack that could offer support for any of the various products that could be included within a converged stack.
VCE took things a step further and actually built a layer of management software around the VBlock to maintain lifecycle, security, and patches/upgrades across all of the components across multiple stacks in one piece of software. Subjective as to whether this is actually desired or not. YMMV. But that’s not what we’re here to talk about.
But what was universal about a converged infrastructure stack? If you strip the brand labels off, you were left with SEPARATE components of infrastructure to handle compute, storage, and networking. Servers, shared storage arrays, and switches all pre-racked, cabled, and configured in a cabinet delivered to your loading dock.
It is a really elegant solution for those that do not have the staff to handle this level of deployment, or do not want to pay the premiums for a VAR/SI to come and implement it themselves.
NetApp and Cisco have dominated the market with FlexPod in recent years, seeing it climb to a $4B market, with VCE and their VBlock being the top two options. FlashStack from Pure Storage also deserves a mention here, as they have been one of the biggest disruptors in primary storage in recent years.
The beauty of this is that you could run almost any workload on it. Virtual infrastructure, Secondary backup targets and long-term archives, VDI implementations, etc. You could also mix and match the models of each of the sets of gear to tailor it to your use-case and needs.
When you hear those words, or the HCI acronym, I simply think of two companies… Nutanix and Simplivity. While others have come and gone, and new vendors enter the market all the time, they have been the stalwarts of the hyperconverged movement over the last 5 years.
In fact, it was at an analyst briefing during the first Nutanix Sales Kickoff with some of the initial folks at Nutanix where the term was coined. This was in 2011. And I have it on good authority that “hyper” had absolutely nothing to do with “hypervisor.”
Unfortunately, due to the nature of their initial deployments being VMware-focused, I believe there has been an inadvertent expectation of some sort of hypervisor involved in a “hyperconverged” solution.
This is, unfortunately, untrue and misaligned. Especially with today’s market seeing a new found resurgence in hyperconvergence.
So, let’s define it!
Hyperconverged is simply “more converged” than converged. Above, I outlined converged infrastructure to be a cobbling together of various components within a single rack, combining the delivery of compute, storage, and networking. With hyperconverged the same concept is applied, but the delivery vehicle is a single chassis with n number of nodes.
That’s it! It’s that simple! These platforms can run VMware, Hyper-V, standalone applications, or simply be storage systems.
Cohesity has introduced the first hyperconverged storage system with a distributed file system focused on consolidating workloads that are often considered “secondary.” When you hear me say the words “Hyperconverged Secondary Storage,” that’s where it comes from.
Hyperconverged Secondary Storage
What if we could take the hyperconverged model of shared-nothing x86 commodity hardware nodes and combine it with a distributed scale-out file system built around some of the fundamentals of object storage to include things like indexing, MapReduce, Elastic Search, and metadata management?
On top of that, we could abstract any number of workloads. That’s Cohesity.
Ultimately, it’s all about consolidation. Much like VMware did for servers, we’ve come completely full circle in that we’re now applying a lot of the same consolidation techniques to data and the systems that manage it. Simply being able to run multiple workloads like data protection, test & dev, as well as home drives and departmental file shares, makes this one of the coolest consolidation platforms to come out in recent years.
Personally, I would have gone nuts for a stack of these! I was running a big datacenter as recently as 2010, before turning to the dark side, and it was a nightmare to manage all of the different pieces of infrastructure across various mgmt UI’s, licensing and support models and contracts…
If you’re interested in knowing more, start with these demo videos here on our YouTube channel.
Yes, it most certainly IS hyperconverged. No question. To me, there is a clear delineation between converged and hyperconverged, and this is backed up by multiple vendors’ products, analyst research, and various industry professionals.
Let me know your thoughts in the comments below!