Convergence is one of those powerful IT concepts that powers progress. Functions that were once seen as separate and distinct get combined and deeply integrated in a useful way, and there’s no going back.
Cell phones meet handheld computers, and smartphones are born. Consumer electronics meets automobiles, and the market shifts. Home networking meets smart appliances, and modern living starts looking like a 1960s science fiction movie.
Convergence plays out powerfully in enterprise IT settings. So many operational models are based on the historical distinct and separate nature of many IT functions. Converge those functions, and the IT model can evolve to a much more powerful and efficient state.
But digging a bit deeper exposes some interesting philosophical questions. For example, how and where should convergence ideally occur?
I’ll give you a hint on where I’m going. I believe the hypervisor is the inevitable platform for infrastructure convergence — both today, and into the future.
A Simple Concept — Or Is It?
Convergence in an enterprise IT setting should be a simple concept, but it turns out to be harder than it looks.
At a high level, you’ve got an individual who needs to get something done: provision a service, monitor a service level, troubleshoot a problem, and so on. Performing that function means touching a lot of different parts of the IT estate.
In an unconverged environment, that means a lot of manual work and human integration. The pieces aren’t designed to work together, and it shows. There’s inconsistency, inefficiency and gaps everywhere. It can be more like herding cats than running smoothly.
But — as any IT practitioner will tell you — simply slapping a slick-looking management layer over disparate components isn’t a satisfying answer.
First and foremost, many IT components aren’t designed to interact intelligently with other components as part of an overall workflow. Storage doesn't know much about compute and network, network doesn't know much about storage and compute, and so on. Interfaces are incomplete, syntax and semantics vary, and so on.
But that's not the only challenge. For one thing, there’s very little standardization in how IT people actually do things on a day-to-day basis.
The all-too-familiar result is a hodgepodge of processes and tools, with people jumping into and out of different interfaces depending on what they’re trying to get done. Looking deeper, you’re struck by the incredible inefficiency of it all — redundancy, incompatibility, incompleteness, etc.
While convergence can’t entirely eliminate this (it’s enterprise IT, after all!), convergence presents a powerful opportunity to tame this jumble of functionality.
Good Convergence Delivers A Better User Experience
In a converged environment, the components are designed to not only work together, but interact intelligently. Everything "knows" about everything else, and acts accordingly.
As a result, the “product experience” is now redefined: you don’t see individual components interacting; instead you get a new meta-product experience that is greater than the sum of its parts.
When I’m using my smartphone, I can’t easily tell when the phone stops and the computer begins — they both interact seamlessly to create a product experience that’s distinctly different than either.
Similarly, in a well-converged IT environment, you don’t experience the traditional boundaries — you’re interacting with a new entity that’s a superset of what came before.
But in most IT settings, that enhanced user experience isn’t just something that's nice to have — it can lead to all sorts of desirable business goals: cost savings, far greater productivity, fewer issues, faster resolution of problems, more efficient use of resources, increased agility and so much more.
Hardware Infrastructure Convergence
I will argue that there are two distinct forms of infrastructure convergence in play today. Both are powerful, and both can work together.
The goal of infrastructure convergence is simple: bring together the three primary disciplines (compute, network, storage) into a single entity — one where the parts are designed to work together, and interact intelligently.
In the enterprise IT space, one of the more notably successful examples of hardware convergence has been VCE. A joint venture between Cisco and EMC (with VMware participation), it delivers an engineered infrastructure product (the VBLOCK) that’s far greater than the sum of its parts.
Part of VCE’s success has been its maniacal focus on the “user experience” — from requirements to design to installation to operations. VCE’s success has not gone unnoticed, and as a result there are now many vendors now vying for the relatively new category of ‘converged hardware infrastructure”.
I would predict that this category will continue to grow and evolve over time. If nothing else, it clearly meets a critical customer requirement, and demand shows no sign of abating.
That being said, I can also make a case that hardware convergence — over time — will give way to software convergence.
Software Infrastructure Convergence
If you fully internalize the industry move to industry-standard server architectures, virtualization and hence implementing in software what was once done in specialized hardware, you can see the attractiveness of software infrastructure convergence.
Indeed, at the core of SDDC (software-defined data center), there’s the premise that all core infrastructure disciplines (compute, network, storage) are not only potentially implemented in software, but also fully converged.
Doing infrastructure convergence in software offers some tantalizing potential:
- the converged environment becomes largely agnostic to the underlying hardware
- the resource pool is both larger and more fungible: a given resource can be dynamically allocated to providing either compute services, network services or storage services as needed vs being statically allocated.
- done well, the infrastructure components are aware of each other, interact intelligently, and play nicely together
- as a result, new and more powerful management integration and automation models become achievable: everything is software, everything is dynamic, everything is orchestratable.
The VMware Angle
When it comes to software infrastructure convergence, there’s only one place that makes sense to do it — and that’s the hypervisor. It’s become the key layer between the application and the infrastructure, and knows a lot about both.
For those that still claim that server virtualization has become a mere commodity, I would argue that server virtualization was just the opening salvo of a far more powerful concept: software converged infrastructure.
It should be obvious that VMware has moved far beyond familiar server virtualization, and now is far out ahead on the frontier of software-based infrastructure convergence. While the new networking and storage capabilities may seem new and unfamiliar to some (NSX and VSAN as examples), they are both solid products that deliver key portions of the broader picture of SDDC and software infrastructure convergence.
However, in enterprise IT, it’s more about evolution than revolution (unless you’re a noisy startup company), so it’s clear that software-based infrastructure convergence will be achieved in most settings gradually over time, with full integration of traditional networking and storage devices.
But you can see where it’s going …
Barriers To Software Convergence Adoption
Quite simply, there are three in my mind.
First, the respective converged disciplines have to meet requirements and do what’s expected. No one is willing to give up on critical requirements just to get to a converged model.
Second, we’ve just begun to scratch the surface on what deep, intelligent integration might mean across the three infrastructure disciplines implemented in software.
Yes, there are interesting examples to point to today, but there’s far more that’s achievable. And it’s this deep, intelligent integration that delivers the experience that makes convergence so compelling.
Finally, the real heavy lifting will be in larger IT shops that have to fundamentally re-engineer their operational processes, skills, roles, etc. around what convergence can achieve. We saw it with the first wave of hardware infrastructure convergence, we’ll see it again with software infrastructure convergence.
Industry Implications Of Infrastructure Convergence
If we go back to consumer electronics, once things converge the buying criteria inevitably shifts. The converged “product” becomes the desired experience; the historical predecessors become less interesting over time.
Shifting back to enterprise IT, it wouldn't be unwise to assume the same sorts of things will happen. What will become more important over time — stand-alone networking, or networking that’s fully converged with compute and storage? Stand-alone storage, or storage that’s fully converged and integrated?
Put differently, I think it’s going to be increasingly difficult to claim “leadership” in any software-defined category (e.g. storage and networking), unless you can show the entire convergence story across all three disciplines — preferably software-based, hardware-agnostic and fully open. The value should quickly shift to how well things play together once minimum functionality requirements are met.
Or, to be more provocative — at what point do networking and storage functions simply become a feature of the hypervisor?
Is The Future Closer Than You Think?
Now that I'm firmly ensconced at VMware, I'm getting a much closer look at what customers are doing with the technology.
There are literally many thousands of IT organizations who buy into the SDDC vision, are deploying the technology and re-engineering their operational models as a result.
It's pretty darn amazing to see what these people are doing.
For them, convergence isn't some abstract concept.
It's how they're building their next-generation IT.
-----------------
Like this post? Why not subscribe via email?
Chuck,
It is obvious to most people in the industry that convergence is relevant today and the way of tomorrow but the separation of software and hardware convergence is an interesting thought. We see solutions like VIPr converging the software layer and solutions like VBlock converging both the hardware layer and software layer but which will be the first adopted? In the consumer electronics world first adoption was software (Windows) but it is the complete convergence of both software and hardware that is winning today (Apple). Will we see the same progression for enterprise IT? If so which is the right direction to take today for a vendor? Can you afford to play the short game of winning SDDC and loose the long game of completely converged infrastructure? Only time will tell but it will be interesting to watch.
Posted by: Henry Johnson | January 08, 2014 at 10:00 AM
Hi Henry
The comparison with consumer electronics is a good one, but I think it has its limits. If we consider smartphones, we can see both models in play today: converged software/hardware (e.g. iPhone) and hardware-agnostic (e.g. Android). Both seem to be doing well.
But enterprises (especially larger ones) are motivated differently than we as consumers. And -- for most -- having more choices is better, all things being equal. Once of those choices could potentially be a converged hardware platform -- independent of the desire for a software-converged environment.
What I didn't get in to here -- but would like to -- was the evolution of standard, commodity server hardware into what looks very much like a converged hardware platform.
Thanks for the comment.
Posted by: Chuck Hollis | January 08, 2014 at 10:33 AM