The individual infrastructure components are straightforward enough when considered individually; getting everything to interoperate smoothly is far more burdensome.
As hardware prices fall — and IT capacity balloons — the focus has clearly shifted to improved operational efficiency: doing more/faster/better but without adding more people. “Simple” is the new killer app in enterprise IT infrastructure.
All three approaches to integrated infrastructure — converged, hyper-converged, hypervisor-converged — have that singular goal in mind: making things simpler.
But each approaches the problem in a different way — and with different results.
With Best Intentions
Most IT infrastructure vendors invest heavily in their products so that they can be used with others: providing external interfaces, testing for interoperability, writing lengthy reference architectures, and so on.
But despite best intentions, the resulting environments end up being notoriously difficult to architect, integrate, support and manage efficiently on a day-to-day basis.
As Philip K. Dick said “exactly what the powers of hell feed on: the best instincts in man” — a purgatory many IT professionals know well.
Form inevitably follows function: look at most traditional IT organizational charts, and you can see familiar server teams, network teams and storage teams. Since it’s all infrastructure, workflows often bounce back-and-forth between the silos like a pachinko machine.
It’s inefficient and it’s wasteful. And just about every IT group is under enormous pressure to move forward -- on to something better.
The problem with using any term in an IT discussion is that definitions are imprecise -- exactly the situation here. There are no standards bodies or arbitrators, so we make do the best we can.
At its essence, convergence is a simple idea: make disparate components act as a whole.
Perhaps the best example of convergence is the modern smartphone, which — in addition to making plain-old voice calls, is also a camera, music player, GPS, web browser, video player, gaming console et. al.
As a result, you can drive amazing workflows thanks to convergence on smartphones.
One example: my daughter took videos of a recent concert, edited it, added her voice-over commentary, and posted it on a half-dozen social sites.
While she was catching a ride back from the show.
Good thing I subsidize her data plan :)
Not A New Idea
Through my jaundiced eyes, convergence in IT infrastructure isn’t really a new idea. During the golden age of minicomputers, these early examples of converged infrastructure came with their own purpose-built hardware, networks, storage, databases, etc.
The good news? Everything usually worked together as you’d expect — and were very operationally efficient as a result. The bad news? Functionality was limited to what the vendor provided, and your choices were always very constrained.
The pendulum swung: UNIX, Windows, client-server — and complexity increased as a result.
Now it appears to be swinging back.
#1 -- Converged Infrastructure
In my eyes, the first modern take on converged infrastructure was the VCE Vblock — something I was a big fan of during its inception. VCE — a separate company — delivers a pre-integrated product — a Vblock — built on technologies from VMware, EMC and Cisco.
NetApp quickly followed with FlexPod — an alternative take that initially focused on a reference architecture: assembled, delivered and supported through partners.
Shortly thereafter: HP, IBM, Hitachi and others followed. EMC offered up VSPEX which allowed more choice on server and network — but using an EMC array, of course.
IDC defines the category as “integrated infrastructure” — as opposed to “integrated platforms” that are designed to run one type of workload (e.g. Oracle, Teradata) vs. many.
This has quickly become a big business.
In Q1 of 2014, IDC says integrated infrastructure approached $2 billion of revenue — a breathtaking 69% of year-over-year revenue growth, not to mention over 650 petabytes of new storage capacity — no wonder the storage vendors are paying attention :)
That's approaching a $10B category that didn't really exist just a few years ago.
All of these converged offerings have a similar premise: integrate familiar infrastructure components to dramatically simplify the sizing, design, acquisition, installation, support (and to a certain degree, management) of IT infrastructure.
However, your choices — while good ones — are necessarily limited. For Vblocks, it’s the vSphere hypervisor and management stack, Cisco servers and networking, with EMC storage and data protection. For FlexPod, the same generally schema, except with FAS arrays. Similar for HP, IBM, Hitachi et. al.
Despite the appearance of the dreaded lock-in, more than a few people are willing to accept a few reasonable restrictions in exchange for a far simpler experience.
#2 -- Hyper-Converged Infrastructure
A few startups in (SimpliVity, Nutanix and just recently NIMBOXX) have taken the idea a bit further — by using software to eliminate the need for external storage. All sell pre-integrated modules that are easy to order, install and — presumably support.
While IDC does not track this category (yet!) —- and the companies involved are not public — it’s hard for me to get a bead on what’s actually getting sold vs. evals, POCs, etc. I can only assume that all are enjoying modest success -- but nowhere near the converged infrastructure category.
Understandably, there are tradeoffs involved.
For each, there’s only one primary storage choice: the software stack that is provided by the vendor — no external storage as you’d find using a hypervisor-based approach. And, of course, you’re limited to the specific server choices that are OEMed by each.
That being said, I’ve seen situations that would be well-served by simply buying a few modules and just getting on with it. And I'm sure we'll see more offerings in this category before too long.
While these solutions are arguably “more converged” than the previous category, calling them “hyper-converged” is a bit of a stretch. Network functions are still largely separately managed, so there’s that. Since two of these offerings use vSphere, you’re presented with at least two (or more) distinct management experiences — the one from the vendor, and the one from the hypervisor.
And there’s only so much you can do with plug-ins :)
To be clear, I’m speaking on my own here — VMware has not widely adopted this term, except when describing VSAN. But -- when used more broadly -- I'm finding it a useful term that accurately describes what makes software-defined data center (or software-defined converged infrastructure?) so compelling.
At a conceptual level, the hypervisor sits in a very strategic position between applications (and their operating systems) — and the underlying infrastructure. It can abstract compute, network and storage. It can provide a powerful control plane over each, as well as additional functionality.
And — because it can abstract all the underlying resources, the hypervisor can converge the operational management experience in a way that’s almost impossible in other ways.
Sure, prior to smartphones, I could have shot a movie on a video camera, transferred it to my computer for editing, recorded a voice-over, and individually uploaded it to multiple sites. But it would have taken many more steps, taken a lot longer, and problems would have inevitably cropped up along the way. Converged is easier.
Convergence — deep integration across normally disparate functionality —- occurs in the hypervisor. As a result, you’re now operationally managing the hypervisor’s abstractions, and not the underlying subsystems.
The resulting workflows can be natural and seamlessly cross boundaries: optimized around form vs. function. It’s an approach to convergence that's designed to be automated — it’s not an afterthought.
To be clear, “hypervisor-converged” doesn’t mean all the functionality is provided by the hypervisor alone -- just that functionality is abstracted -- although software-based functionality will certainly be an option for some.
For example, VMware’s NSX can effectively exploit underlying functionality on a wide range of networking gear, including Cisco’s. When vSphere VVols (virtual volumes) become generally available (now in public beta), the same can be said for external storage arrays. And, of course, vSphere has been exploiting physical server hardware for a long time, and does it well.
Contrasted against other forms of convergence, one thing is obvious — a hypervisor-converged environment can run on just about any reasonable choice of gear an IT shop wants: server, network and storage. That's attractive to many. But at the same time, all those choices can create complexity, especially with regards to sizing, procurement, integration, support etc.
We’ve already seen some “appliance-ization” of hypervisor-converged infrastructure (e.g. VSAN Ready Nodes), and I think it’s safe to say we’ll see a lot more before too long.
A hypervisor-converged solution that can either be consumed on an appliance-ized module, or run side-by-side with roll-your-own hardware will be very, very attractive to many, I think.
More Open Than You Might Think
And then there’s the concern that some express regarding potential hypervisor lock-in; specifically vSphere as it's so widely used. Note that the significant majority of converged and hyper-converged products are already built on vSphere today. Since the workloads themselves are eminently portable to other hypervisors and public clouds, it's really the operational model that deserves the focus.
Although I think there’s far more choice and openness here than generally recognized.
As a starting point, it’s long been the case that there’s a rich choice of management tools and third-pary add-ons -- in addition to the ones that vSphere provides as part of vCloud Suite. Arguably far more choices than any other hypervisor. Not to forget, the VMware-provided management tools are quite adept at managing non-vSphere environments.
What’s more recent is the great out-of-the-box integration today with OpenStack. As an example, check out this eye-opening video showing mix-and-match interoperability of NSX and vSphere with KVM-based hypervisors managed by Openstack Horizon. That's right: mix-and-match hypervisors -- although not deep integration, which is required for convergence.
Certainly more to come here.
Convergence Is Here To Stay — And Grow
Maybe I’m biased, but I see the evolution here as a logical progression: converged, hyper-converged and ultimately hypervisor-converged.
All important steps towards the end goal of simplicity.
Maybe this whole software-defined thing is the real deal?
Like this post? Why not subscribe via e-mail?