Converged infrastructure and hyper-converged infrastructure have been all the rage for the last few years. Most major vendors are in it including “independent” ones like Nutanix, SimpliVity, among others. Their promise is simplicity: easy to deploy, easy to configure, easy to manage. Certainly it sounds appealing and many buyers appear to be opening their wallets.
However there seems to be some confusion around what the words exactly mean. The consensus looks like converged systems represent discrete compute, storage, and networking sold as a single SKU or on the basis of a reference architecture with all the components guaranteed to talk to each other. Whereas hyper-converged systems are about pre-integrated pieces in which everything is tightly connected and parts cannot be sold separately, an appliance form factor, with a management layer on top. A lego brick like building block approach. Though being preconfigured does remove some flexibility. Smaller blocks could perhaps help, a need that some of the vendors appear to be meeting.
Confusion is probably being caused because the terms are primarily vendor driven. No standards bodies specifying what convergence or hyper-convergence mean, which does sound like potential for significant interoperability challenges down the road. Not surprising then that vendor lock-in is frequently mentioned as an outcome.
Let us step back: The concept is not really new. There has been talk in the past of modular data centers, data center in a box, cloud in a box, appliance, and what have you. Nevertheless the vendor-driven hype (pun unintended) around converged infrastructure and hyper converged infrastructure seems to have reached a crescendo. Possibly some of it has to do with the move to software defined networking and software defined storage (an overused term in and of itself) based on commodity hardware, software switches and routers, white box components, open source, and so on. Clearly SDN (and SDS in its “classic” sense) hold the promise of significantly lower costs, hugely greater flexibility, and more importantly control over the infrastructure. From the customer’s perspective, it is all positive provided they have the skills to put it together.
Yet from the infra vendors’ point of view, SDN and SDS probably do not look that rosy. Having the buyer in control is likely not a good idea. Revenue negative or neutral at best. Long term, SDN definitely appears to represent a threat. So figuring out a new way to sell boxes would be critical. And converged and hyper converged infrastructure certainly seem to be a smart way to counter the SDN threat, create a wave around something, and drive revenue growth. Selling boxes the old way no longer looks to be applicable. If a new future is inevitable certainly it makes sense to try and control it.
That customers are looking for increased agility and ways to manage their project specific workloads where everything is business driven, application driven is the message. Putting the app on top is what SDN is all about. So the hyper-converged vendors appear to have gotten onto that bus with the claim that their solutions are what is needed to align infrastructure with applications. They seem to have co-opted the “software defined” messaging.
An aside. There has been talk about how 70% of IT spend is “keep the lights on” and how converged infrastructure will change that dynamic, raise spending on innovation. Perhaps spending on converged infrastructure boxes is a proxy for innovation spend!
To digress a bit, parallels appear to exist between the move to converged systems and the earlier shift from custom software to standard, packaged software. Pre-built functionality, the promise of faster deployment, on one side with the loss of flexibility and control on the other. Flexibility and change was a challenge with standard software and the same story seems to be playing out here. Another familiar issue is the whole thing of people buying more than what they need: excess functionality in the case of software and additional boxes in the case of converged infrastructure. Buying too much hardware could have an environmental impact but that is another story.
End of the day, it should boil down to the little matter of control. Software systems that can drive any choice of storage hardware at the backend, putting buyers in control. Open source software options would be even better. Bring your own storage is what it will be. Power to the buyer: The seesaw shall tip to the customer, the “new master.”