A Microsoft executive has thrown cold water on the hype around virtualisation today, even as the company seeks to promote Live Migration features released as part of Windows Server R2.
Program manager of core virtualisation at Microsoft, Ben Armstrong, said there were two 'cons' being neglected in commentary on virtualisation technology so far.
"The two biggest cons are, no matter what anyone will ever tell you, performance server overhead to virtualisation -- there has to be,” Armstrong said in an interview with iTnews at TechEd on the Gold Coast. "We worked very hard to make that as small as possible, but you will always get better performance running on physical hardware."
Another con, Armstrong said, is the “complexity tax” that comes with a virtualised environment.
"[If you had a] bunch of physical computers and you’re managing them and move those into virtual machines, you reduce the amount of hardware you have," he said. "You’ve now added an extra layer in the stack that you need to manage."
Armstong said that Live Migration in Hyper-V worked similarly to VMware’s VMotion product.
"[It's] where you can take a virtual machine from one server and move it to another physical server with no noticeable downtime on the network. [It's] fantastic for people who are needing to do systems maintenance," he said.
"Let's say you’ve got a couple of Hyper-V servers running a couple of virtual machines and you need to change the network card or you need to do something on the physical box," he said. "Traditionally, you’d have to wait until after hours to do that. Now you can go in - in the middle of the day - and move the virtual machines off, do your maintenance, and move the virtual machines back. No one notices that it happened."
But Armstrong said customers are quite rightly hesitant to "put all their eggs in one basket."
"Users tell us they’ve got 50 servers," he said. "[They'd say] any one of them [being down] was not critical. Their DHCP server could be down for half an hour, whatever. Their DNS server could be down for half an hour, whatever. But to take those 50 servers and stick them on a physical computer and say that hardware failure is going to cause the outage of my entire server infrastructure? Yes, it’s catastrophic."
Clustering, he said, was the answer to the problem.
"What we’ve worked very hard on is that Hyper-V works with Windows failover clustering to allow you to actually create clustered instances of Hyper V where the virtual machines are treated as highly available objects," he said. "If you had a hardware outage, virtual machines get moved straight across to running hardware.”