If you had to identify a mega-trend in the server virtualization world for 2012, what would it be? Predicting the future is never easy (hey, where are all the flying cars they promised us?), but Oded Haner, CTO at virtualization management software maker HotLink, reckons he has spotted not one but three. (Mega-trends, that is, not flying cars.)
Mega-trend No. 1 is the adoption of multiple hypervisors in organizations. Haner reckons it's because the alternatives to VMware are now so good that companies will adopt virtualization technology according to the applications they want to run and the in-house skills they already possess. "Companies with a Microsoft background will run Exchange on Hyper-V; those using XenDesktop will use XenServer, and companies using JBoss will adopt Red Hat's KVM. The process of adoption is simple, as they will need no new skills," he said.
Haner said that he has seen plenty of companies running proof-of-concepts this year. They have discovered that running multiple hypervisors is not nearly as complex as they anticipated. "There is money to be saved, and if there is an ROI in adopting multiple hypervisors and little risk, then there will be a tendency for that to happen," Haner added.
Mega-trend No. 2 is a change in focus from the hypervisor itself to the server virtualization management layer. What Haner is talking about here is the problem of how you carry out virtualization management if, as mega-trend No. 1 predicts, adoption of multiple hypervisors becomes commonplace. The problem, as Haner sees it, is that if you try to use Microsoft's System Center to manage vSphere, for example, or perhaps XenServer to manage Hyper-V, what you'll be able to do is carry out basic tasks on the hypervisor but not the common (though not quite day-to-day) stuff. "You end up having to use a second management layer to do that, so you need to have two sets of skills and you have to pay for two management layers. That approach is bound not to be successful," he said.
HotLink and other start-ups are trying to provide a kind of translation layer so you'll be able to use vCenter, for example, to manage Hyper-V, Xen and KVM "natively." The translation layer will take vCenter commands and translate them into a form that works on the other hypervisors. "What's needed is a way to decouple management from the hypervisor," Haner explained.
You'd imagine VMware especially would be reluctant to let this happen--after all, anything that makes it easier to use multiple hypervisors makes it easier for customers to move away from the VMware ecosystem, or to use a VMware hypervisor without paying for vCenter by managing it with something else. The same is probably true of all the other server virtualization platform makers. But in fact VMware and the rest of the hypervisor merchants are being pretty cooperative, Haner said.
Mega-trend #3 is an increased opposition toward cloud lock-in. It's the old Roach Motel scenario: You implement a public cloud or a hybrid cloud and then find you're a bit stuck when it comes to moving to a new service provider. More generally, what if you're using, say, VMware's vCloud technology and then decide you want to move to a cloud service provider that uses OpenStack or Amazon's AWS API?
One answer to that is simply this: Isn't competition between vCloud providers or OpenStack providers enough to guarantee that you are not locked in to a particular provider, even if you are locked in to a particular platform?
Haner doesn't think so. He said that indications from customers he talks to suggest they really will want the ability to move between platforms, and therefore they want technology that allows them to migrate workloads from, for example, a vCloud environment to an OpenStack one. And, of course, companies like Hotlink and others will be happy to oblige, with cloud gateways that carry out translation in an analogous way to the translation that's required to manage hypervisors using the "wrong" management platform. That's the plan, anyway. "We hope to offer transformation technology that transforms virtual machines and the metadata that goes with them, and is able to register the workloads on to other platforms," Haner explained.
All three of these mega-trends have an air of plausibility about them. Whether they actually happen in 2012 depends to a large extent on whether the sorts of products that HotLink and others want to bring to the market end up being released on time and doing what they promise. They sound great, but you can't help wondering if they'll take much longer to come to market than some people are predicting.
Much like those flying cars, really.
Paul Rubens is a journalist based in Marlow on Thames, England. He has been programming, tinkering and generally sitting in front of computer screens since his first encounter with a DEC PDP-11 in 1979.