VMware’s vSPhere: “Bigger, More Open, More Resilient”

VMware’s vSphere announcement had been widely anticipated, but the company still made a point of playing it big. As well as drawing the CEOs of Dell, EMC and Cisco to an event in the US, the company ran preview briefings elsewhere.

Why is this such a big deal, we asked Reza Malekzadeh, senior director of marketing for EMEA, at the London event. vSphere 4 is really just a new version of VMware’s Virtual Infrastructure (VI) product, isn’t it?

“vSphere is much more,” he said. “It’s a cloud OS. Where a computer operating system manages the screen, keyboard, disk and other hardware, vSphere delivers the same thing for a data centre. Not just servers, but storage and networking.”

This is a step beyond other virtualisation vendors, he said, because “Microsoft and Citrix only offer server consolidation. They let you pool virtual machines.” VMware has made this next step because it has a five year lead on the others, he went on.

Maybe so, we asked him, but it could be argued that VMware’s history leaves it tied to an older way of doing things. The Xen hypervisor used by Citrix is based on open source code (and Microsoft’s Hyper-V is closely related). Those companies would argue that they started with a newer model, built for hardware that is increasingly designed for virtualisation, so they can be more flexible.

Open source … or just open?

“It’s true that Citrix is an open source technology,” admitted Malekzadeh, conceding that this has benefits. “VMware is not open source, but it is an open technology. You can switch from us to Microsoft or Citrix for free.” This is because VMware has published its formats and there are conversion tools, he explained. “All our management work is adopted by the DMTF standards group, and our cloud APIs are published to standards bodies.”

“If you want to move to Citrix or Microsoft, you are free to do so,” he said. “You stay because you get value from VMware.”

Isn’t it likely that open source products will be cheaper, we asked. “That’s a misconception,” said Malekzadeh. “It can be cheaper up front, but are you getting the value you expect?”

If VMware is not giving value, he asked, why is it still growing: “We did $500 million in virtualisation in Q4 of 2008,” he boasted. “Citrix did $9 million.”

Citrix’ argument has been that VMware’s business model forces it to move towards selling more management tools, creating a caste of VMware administrators, and that much of its revenue comes from selling management tools, when data centre staff might prefer to manage their virtual servers with existing general purpose tools. Malekzadeh does not see this as a problem: “vSphere has the same management console we have always had. If we are not giving value, people wouldn’t buy it.”

Closer ties with Cisco?

What about the network part of this, we asked. Was analyst Roy Illsley right to see this as the main part of the announcement?

“It depends who you are,” said Malekzadeh. “If you are coming from a storage background, you would be impressed by that part.”

However, the notion of moving virtual machines round a network has given the industry something new, he said. It used to be that virtual machines broke when they were moved across a network, he said: “Our partnership with Cisco fixes that. That is very critical.”

If it is that important, and it comes at a time when Cisco itself is moving into servers, can we expect a closer relationship between the companies in future? “No comment,” he said.

He also declined to comment on whether VMware would extend its partnerships to other network vendors, such as Juniper or Extreme – but he explained that the partnership is not as exclusive as some might claim.

“Our distributed switch will work with any vendor’s hardware,” he said. “We don’t care what switch vendors you have.” The vNetwork part of vSphere does allow a data centre manager to move virtual machines across network switches from multiple vendors, he said.

What the distributed switch does, is it allows VMware users to manage the network using Cisco’s native commands. That’s not something that could be offered with other switch vendors, he said.

Performance and resilience

Finally, we touched on fault-tolerance and the performance part of the story.

VMware really is pitching its fault-tolerant feature squarely against the traditional fault tolerant vendors, which still carry on, from the heritage of Tandem and others, said Malekzadeh. They used specialist hardware, but VMware simply links virtual machines on different servers.

The full fault-tolerance in the new vSphere does cost extra however. The previous version of the product allows one virtual machine to shadow another, but with some latency: “If your virtual machine crashed, you might lose a couple of seconds of processing.”

For some users, that’s good enough. For those that want more certainty, vSphere now allows two virtual machines to operate in lockstep: “It’s more expensive, but some users require that level of resilience.”

Finally, the new vSphere allows massive scalability, he said, with 32 physical servers combined into one virtual computer, and large numbers of these resources can themselves be combined.

“The real message today is we can handle pretty much anything.”