Sysadmin blog The virtualisation community is all atwitter with speculation about VMware's MARVIN. MARVIN, for those who don't know, is supposedly VMware's hardware'n'software answer to hyper-converged competitors like Nutanix.
Everyone has an opinion, and everyone has a theory – but VMware itself is carefully keeping schtum. I have some hypotheses of my own.
Multiple sources at VMware have told me that MARVIN – at least in its first iteration – isn't going for the enterprise. It will not leap off the stage at VMworld 2014 and roundhouse kick VCE or even Nutanix in the brain.
MARVIN, stage one, is targeted at the "small to medium" IT shop. Given the strength of VMware's portfolio, I'd normally say that this is where everyone from small fry like Synology and Western Digital through to volume players like Dell and HP should start pissing themselves in terror. In truth, however, they're probably safe too, if only because VMware is institutionally incapable of grokking the "small to medium" market.
The icing on the cake is that, within the halls of VMware's Palo Alto headquarters, the tidbit that is most likely at the core of MARVIN – VSAN – is the political equivalent of a live wire. So political has it become, in fact, that one of my contacts simply refers to discussions about VSAN as "global thermonuclear war". The only way to win, apparently, is not to play.
All of this leads me to believe a few things about MARVIN. The first is that it is going to be awesome; a demonstration of just what is possible when you start putting VMware's technologies together in one package.
Secondly, I think it is going to end up crippled in some fashion that doesn't make rational sense, unless you start incorporating the complex political interactions that companies within the EMC federation are burdened with. Lastly, it's probably not going to priced appropriately for the target market, at least not on the first try.
What MARVIN will be is a technology demonstrator. The first concept car of the upcoming software-defined infrastructure wars. VMware wants to be first here, not only because they have the technology that means they actually deserve to plant their flag, but because by being first they get to name things.
If VCE was "converged" and Nutanix or Simplivity are "hyper-converged" VMware wants you to think of MARVIN as "hypervisor converged" or, at the very least, "the software-defined datacenter." I'm honestly not sure if they've finished hashing out the marketing lingo here – I'm putting together bits and pieces from all over – but suffice it to say that real money is going to be pumped into convincing us all to call software-defined infrastructure something that evokes thoughts of VMware.
But just why, exactly is MARVIN likely to be amazing, and what do I mean when I talk about "the upcoming software-defined infrastructure wars"? To understand this fully we need to dive into what VMware provides as a complete stack, and why a super-secret project like MARVIN was needed in the first place.
More than Nutanix
MARVIN pretty much has to be more than just what Nutanix brings to the table. There are a number of VSAN ready nodes that have already passed inspection and received VMware's official blessing. It doesn't take a secret project to build a management interface that doesn't suck for cluster management, and VMware honestly believes that by building VSAN into vSphere they've got a UI that can go toe-to-toe with Nutanix.
Logic dictates that VMware will be bringing more to the MARVIN party than just VSAN. Picture a Supermicro FatTwin with ESXi installed, and a raft of VMware software pre-loaded. Not just VSAN, but vSphere Data Protection Advanced, vCenter Site Recovery Manager, vCOPS, vCloud Suite and NSX. Include vCloud Hybrid Service in the config so that migration or backup of workloads to VMware's cloud is push-button simple, and even roll out a model or two that are pre-canned VDI nodes with all the trimmings.
We're not just talking here about a Nutanix competitor. Every tin shifter on Earth will be moving VMware VSAN-based Nutanix competitors in a matter of months. We're talking about the first honest-to-goodness infrastructure-as-a-service appliance. Simplivity will quibble – they include things like WAN accelerators and a few other goodies that VMware may not – but the truth of the matter is that the full VMware stack can trounce any competitor on the market, even Microsoft.
If you want to play "roll your own", the awesomeness of the gear means absolutely nothing. Even VMware thinks NSX is difficult to configure and thus sell, and from experience, integrating the whole stack of VMware products is "making System Center behaving in a rational fashion"-class painful.
With some of these features – like NSX – having gone live just recently, competing providers (cough, cough, Simplivity) probably won't get to be the first company to truly earn the title "Supplier of Software Defined Infrastructure". Barring a dark horse in the next two months, MARVIN will see VMware first to the top of the hill.
If you're just flogging servers or arrays, bail out now
Make no mistake: this is the game that is being played today. If all you shift is servers, you're dead. If all you shift is storage arrays, you're dead. Server SANs – like VMware's VSAN, Maxta, Nutanix, SimpliVity and so forth – are absolutely going to be part of the solution, but they are not on their own enough to win the war.
Cloud computing in all its various forms has introduced us to a world where you don't have to handhold your infrastructure. Arguments about OpEx are swaying multi-billion dollar enterprises away from owning their own gear and into Amazon's black hole of lock-in doom.
You don't counter that kind of convenience and simplicity with Yet Another SAN. Amazon, Microsoft, Google et al are a threat to every tin shifter and packaged software vendor on the planet. Do you really think VMware is going to eventually give up on hypervisors, make Yet Another Mobile App and hand Microsoft its 30 per cent on a silver platter?
Not gonna happen.
So vendors like VMware – and Supermicro, Dell, HP, Lenovo, etc – all have to adapt. They need to be able to provide infrastructure that is as easy to use as the public cloud, but offers all the benefits of ownership.
That means a truly converged infrastructure play. No pissing around with System Center Designed For People Who Keep Their Pee In Jars Sanity Decompilation Manager. No screaming obscenities at ESXi because it doesn't quite work with your IPMI controller or the hardware sensors in your server. No laborious setting up of monitoring or backups or puzzling over API integration of something or other.
Cloud in a box. Open box, plug in the cluster, turn on, and receive "cloud".
An out-of-box experience complete with all the trimmings, that doesn't cost $RAND*2. If I've money to bet, that's what MARVIN is to be. By this time next year it won't be the only one. The software defined infrastructure wars are only just beginning. ®