How Multiple Hypervisors Might Actually Save You Money

Diversity brings complexity. If you want to maximise the benefits of any IT platform or service and reduce the costs, it helps to standardise on a single choice. But when it comes to virtualisation, the odds are growing that you'll end up with more than one hypervisor to manage — and that might work out cheaper.

Pizza box picture from Shutterstock

That's certainly proved to be the case at Deakin University. Deakin embarked on a major virtualisation project in 2009. With VMware offering enterprise site licences to the entire university sector, it seemed likely Deakin would standardise on using its products for virtualisation. But it didn't quite work out that way.

"In our data centre in 2009, all of our focus was on quality," Craig Warren, infrastructure services director at Deakin, told a press briefing at Cisco Live! 2014 in Melbourne today. "Uptime was the only word that really mattered. Agility was starting to become a focus, but overall we needed to have uptime being premium."

"We had 350 pizza box servers at that stage, and we were going to our vendor and buying a six-pack of more every month. We knew new clients were coming through the door so regularly we might as well order extras because we knew we'd need then."

Virtualisation offered the potential to reduce those numbers, and Deakin wanted to standardise on a platform. "When you have a lot of IT people in a room, you have a diversity of opinion," Warren noted. "The main argument was over whether to go with VMware or Xen; Hyper-V really wasn't an option at that stage," Warren said. "The enterprise site licence kickstarted our virtualisation journey."

Five years later, 96 per cent of Deakin's systems are virtualised and it runs 1200 virtual machines. "We reduced in our DC from 11 racks down to four," Warren said. "We now have only 88 server blades in our DC providing those 1200 VMs and we think there's a capacity within that infrastructure to provide up to 1600."

That presents Warren with an unusual but welcome dilemma: explaining to the university's CFO why there is excess capacity available. "It's not a problem I've had to deal with before, but it's nice having a little bit of spare capacity there."

Provisioning times have also fallen. "With the pizza box environment, it still took us eight weeks from when a customer wanted an app to when we provisioned it," Warren said. "These days with our virtualised environment we can provision a virtual server in under 10 minutes with a couple of clicks." (Deakin uses an orchestration tool it built in-house.)

Patching is simplified too. "Twice per annum we patch all the servers in our private cloud," Warren said. "It used to take 15 people four weeks. Patching periods meant four weekends in a row with nothing. Customers noticed that and they hated it. There's someone logged in at every hour of every day."

"Now with 1200 VMs, it takes two people a weekend to watch the patching go through — it's all automated. We recently updated 28 of our blades, and we upgraded 700 VMs over three days with zero downtime."

"These days we're not buying a six pack of servers a month, we're adding 20-30 VMs a month to our environment. What used to take two months now takes less than 20 minutes."

Despite all those benefits, the virtualisation project has run into snags that mean standardising on a single hypervisor is no longer realistic for Deakin.

"We've had an organisational strategy to be as homogeneous as possible with technology. We have 8500 PCs, they're all Lenovo and they're all black. All our infrastructure is UCS; no smattering of HP or Dell or IBM, and all our switching infrastructure is Cisco as well. So with the enterprise site licence, we thought 'we'll be VMware for life'. That means getting people trained up in just one hypervisor ."

"But VMware has made that hard. They've taken their advantageous position in the marketplace and continued to increase the cost."

At the same time, the alternative options have improved, and the fact that Hyper-V is bundled free with Windows Server 2012 makes it a much more viable option, Warren said. "There could be a $100,000 saving for us by using an alternative hypervisor." While that will require some additional management tools since VMware will also remain in place for much of the infrastructure, the savings are likely to make that a worthwhile option.

Disclosure: Angus Kidman travelled to Melbourne as a guest of Cisco.


Comments

    of course, $100k is the yearly wage of one and a bit sysadmins. If swapping hypervisors increases your need for manpower even slightly, all the savings go out the window.

    (on the other hand if it reduces manual work a bit, FULL STEAM AHEAD)

      Wouldn't you love to be the only sysadmin? Never able to take time off, or get sick?

      There are practical benefits to having a second [or third] person to share the load.

        I agree entirely, and from the numbers in the article it looks like they have more than a few people to share the load.

        My comment was more about how seemingly-large savings can quickly be outweighed by wages in some situations.

          We actually have more than a dozen Sysadmins. They look after our server, storage and data cents as well as all our operating systems. A great team of people. Very skilled.

        same position a friend of mine is in.
        if he gets sick, the whole place can shut down, as he is the only one that knows the systems.
        managemnent have known, and he has asked repeadly for MONTHS to get someone else, management say, we know this, but they haven't done anything.

        where i am, there are 4 of us, one is away for a day, and no big deal.

    VMware's increasing costs are making me look elsewhere as well ...

Join the discussion!

Trending Stories Right Now