All hyper-converged solutions are not created equal

01.04.2016
I’ve just completed a review of hyper-converged solutions (in essence a software-defined system with tightly integrated storage, networking and compute resources) and spoken with a number of companies that have either deployed or are trying to deploy them and it is very clear there is little in common with the solutions offered from individual enterprise class vendors.  

Even the content of the solutions isn’t very consistent. I’m not going to call out the differences by vendor because what is true today will most certainly not be true tomorrow. However, I expect, the core differences between solutions will move between the vendors as each of their solutions matures and the market decides which approach it wants to take.  

[ Related: The pros and cons of hyper-converged solutions ]

I’m going to break them down into four categories: marketing solution, stability optimized, performance optimized and dynamic.

Let’s start by discussing what a hyper-converged solution is.

What hyper-converged means is basically an IT shop in a box, a glass house appliance. The advantages to this approach, if you can do it and it is done right (more on this later), is that everything is designed to work together, complexity is minimized, reliability and ease of management optimized, and both operating cost and implementation time are massively reduced.  

To get to an ideal solution there has to be a high level of collaboration between each of the key components which include servers, storage, networking, (and, in some rare cases, telephony). Like any system the stronger the relationship between the hardware and software components the greater the targeted benefits.

As noted above not all systems are created equal.

These come about when an organization realizes it needs something in market but doesn’t have the time to do it right. As a result, they throw components at the problem and then wrap it with slick marketing to make it look like they have something that is competitive, but it isn’t. I often wonder why some vendors that to do this aren’t sued more often because the resulting offering isn’t anywhere near as good as they represent.

There is a big difference between a mess of aging non-optimized components and a true solution with the core problem that virtually none of the anticipated benefits other than a single vendor puts the mess together are realized.   It is because of “solutions,” and I use this word almost in jest here, that you absolutely have to look under the covers and talk to others who have deployed the platform you are considering before you finalize your choice.

These tend to focus on solutions that mostly come from one vendor. They are heavily vetted and, as a result, give up cutting edge performance for a far more integrated and well tested result. While they trail the performance curve they tend to enjoy the highest customer satisfaction and the greatest level of customer advocacy. Development teams work closely together and a great deal of the product development is significantly connected early in the process.

These tend to be solutions that largely include components from multiple vendors. More of a best of breed approach.   Advantages are they’re more likely to be cutting edge and there are entire industries that need performance so greatly they are willing to trade off speed of implementation and some reliability in order to get it.

The best solutions are created between tightly partnered and often redundant vendors tied to the solutions provider who can choose between them depending on what advantages each vendor has and how close the result is to the customers’ requirements. Advantages are these systems meet high minimum performance standards that can’t be met by stability optimized solutions. While both stability and performance solutions can use GPU components for speed, the performance optimized solutions are more likely to use them aggressively to meet the very high performance requirements of this class.

I expect to see this emerge in the future but, at the time of this writing, I’m unaware of an offering like this. This is where you have a vendor who can blend the stability and performance approaches to provide a range of solutions from a single vendor that doesn’t yet exist but I believe is coming. There really is no sustaining reason why a vendor already capable of building a stability optimized solution couldn’t partner to create a performance optimized offering as well.

And with several vendors having either done massive acquisitions or in the process of doing them their ability to move from a performance model to embracing the stability model is increasing. Once this model takes off we’ll have a clear fight for sustainable market dominance. For now, the opportunity remains divided between those that focus on stability and those that focus on performance.  

First steps are to understand what type of customer you are, as this will significantly change who you have bid on your project. Once you have made your initial choice you need to weed out the providers who are long on marketing and sales promises, but short on having a competitive solution. There is a surprising amount of crap in this highly mission critical segment at the moment and I expect that will remain a constant for some time.

The effort is worth it though, because the firms that have successfully deployed hyper-converged solutions rave about them and it is almost like a religious rebirth. The speed of deployment, reliability and shear lack of aggravation is massively different in a hyper-converged shop and, though you wouldn’t expect it, the agility is actually amazingly pronounced allowing these shops to better compete with offerings from some of the larger cloud providers. In fact, at scale, many have reported they can now offer similar user benefits to a cloud provider at a lower cost to the firm.

In the end, the hyper-converged approach can provide huge benefits, but only if you aren’t tricked into buying a solution that only exists in a marketing executive’s head.

(www.cio.com)

Rob Enderle

Zur Startseite