Most business infrastructure is purchased on an ad hoc basis, cobbled together over time as the organisation buys new tech to deal with each new challenge. This not only removes agility, but also creates data silos.
This is a familiar story for Ryan Kennedy, manager of the academic computing team at the University of Reading. In a recent web seminar (now available on demand), he told us that this approach can also have a huge impact on scalability:
“We used to buy a bit here, a bit there, bolt it on and hope for the best. When you’re looking at doing a project at major scale it just can’t happen… That’s normally the point that you find out none of it’s compatible with each other.”
Kennedy was a websem guest with Seb Hakiel, enterprise architect at hyper-converged infrastructure (HCI) expert Nutanix, which Kennedy chose as the technology vendor to solve the University’s problem:
“We gathered our requirements, went to the University, pitched the idea and proved through a proof of concept that it would scale and support us in the future. We did a small implementation and we’ve doubled it in the last three-to-six months, then doubled it again. We proved that we could start small and scale.”
The scalability of the HCI system was put to the test when the University’s existing legacy storage vendor suddenly went into administration, turning everything into end of life.
“We migrated fully to Nutanix over the weekend and added 400TB in 10 minutes,” said Kennedy.
Elaborating after the websem, he told us: “We had a 15 node cluster…running on [the support vendor’s] servers, and we had a dodgy batch of disks that were in one of them… On one of the occasions when we contacted them we got an automated out of office to say that the company no longer exists… So that is what really forced our hand.”
Rather than dealing with the huge cost of transferring a petabyte of raw storage to the public cloud, Kennedy’s team decided to expand its then-trial with Nutanix. “We use Dell XE hardware, so Dell expedited [10 nodes] and they were with us the week after, racked and stacked and added to the cluster. Once they’re all racked in, you click the expand button and then 10 minutes the later the little graph goes up. It was the biggest R-sink you’ve ever seen! Our customers didn’t notice anything.”
Even the Nutanix team managing the University’s account were unaware of the expansion: “Our account team found out that it had happened after it had been installed, because it was that easy… I don’t know if we still hold the record, but for this we were the fastest reorder ever in Nutanix history. It was 20 days from our original decision to do an evaluation of five nodes to ‘We need more’… It really was just as fast as slotting them in, powering them up and we were away.”
Kennedy and Hakiel joined Computing News Editor Graeme Burton and Technology Analyst Peter Gothard to discuss infrastructure agility; the cloud; and DevOps. Watch the websem now to hear their thoughts.
Save this article