The University of Canberra last year cut by 36 per cent its physical infrastructure footprint as it continues to progressively replace a conventional server and SAN setup with hyperconverged infrastructure.
The process began in September 2013 with the pilot deployment of a Nutanix cluster for dev and test workloads, according to Justin Mason, ITM operations manager at the university.
“We had siloed servers — standalone servers that we would run virtualized workloads on. What we wanted was to move to a hyperconverged platform where all the resources could be pooled,” Mason said.
“So instead of having separate servers to run different workloads, what we liked when we started looking at the hyperconverged options was having a pool of resources, where as you needed more resources you add a new node in; there are no resources wasted.”
The decision to deploy a Nutanix cluster was partly motivated because of its support for the Microsoft Hyper-V hypervisor. The university has since moved to a mixed environment that includes VMware and is looking at standardising on Nutanix’s built-in Acropolis hypervisor.
“Upon completion of the pilot we purchased a second cluster in November 2013 to be used to host production workloads, then in July 2014 we procured a third cluster for our disaster recovery requirements,” Mason said.
“It’s been a gradual progression — there was no overarching plan to replace all of the Dell hardware. What we’ve done has been through natural attrition. As particular infrastructure goes end of life, we migrate whatever is on it to Nutanix.”
“I think the biggest advantage is not only being able to go to a hyperconverged platform, but the shrinking of our data centres,” he added. The university last year decommissioned 40 physical servers.
“We’ve seen dramatic decrease in how much hardware we need in the data centres, which obviously has a flow-on effect in savings on electricity and cooling,” Mason said.
That attrition process is likely to continue, with Mason expecting about 90 per cent of his workloads to end up on the hyperconverged platform.
“There will be some siloed systems that might run standalone servers; there will be some special use cases,” he said.The university recently purchased a new cluster for its Oracle dev-test environments. That cluster will be the first to use the Acropolis hypervisor; if that’s successful then the university will migrate its Oracle production environment next year.