Victoria Uni sheds staff in cloud move

Rolls out elastic infrastructure.

Victoria University expects to shed more than half of its internal IT staff over the next three years as part of continued moves toward a hybrid cloud platform.

The Melbourne-based institution last year deployed two new, automated data centres in an active-active configuration, creating a private cloud platform on which IT staff could easily and quickly provision new infrastructure.

The university significantly consolidated resources – including replacing approximately 350 servers with about five data racks – and achieved a power usage effectiveness rating of as low as 1.2, saving 80 percent of its data centre costs.

Victoria University's executive director of technology and knowledge strategy Phil County told a data centre conference last week that the university's move to consolidate data centres was part of an effort to make it "cloud ready".

"I intend that that data centre is the last data centre that we will ever build," he said.

"Any other data centre deployments in the future, I hope, will be through infrastructure as a service or some other type of service in the cloud."

The university had provisioned some 390 servers in the new data centres over a 12-month period, 90 percent of which were virtualised.

But County said the data centre move has had an immense impact on staff, who had to shift thinking from "thinking about the metal" to focusing on software and underlying architecture instead.

He said he sought to instill in staff a sense of infrastructure becoming 'technology as a service' – "it doesn't matter where the service is, as long as you can get access to the service you want".

"Our data centres used to be almost used like offices; there'd be people wandering around inside, they'd be working there," he said.

"The new data centre model should be dark and automated, because once it's dark and automated in a behavioural sense, it can be anywhere."

Shedding staff

County noted that a single IT staffer now took care of the physical operational aspects of both data centres.

Two-thirds of the university's systems engineering staff left the organisation in a major restructure.

Up to three-quarters of the remaining departmental staff could be shed from the organisation over the next three years as a result of the continued moves, County added.

"We already are reducing our numbers quite rapidly," he said.

"Ignoring the actual parts which are always going to be specific to a university like a research project, we're seeing it already where a lot of the infrastructure services, we're getting them from outside.

"We've gotten to the point now where we've started to look at moving that technology as a service out of the organisation, we're now getting into a touchy point."

County said Victoria University had worked with the unions throughout the restructure to allow for retraining of misplaced staff and redundancy programs where available.

The university had begun experimenting with a hybrid mix of the university's private cloud and external IaaS providers but County said significant moves in that direction were still some years away.

Future models could include linking Victoria University with resource pools hosted at other Australian universities, as well as straight public or public cloud solutions from other providers.

"I think within ten years, the technology is going to go back out to the business units – the choice of the technology will be based on the service and where it comes from doesn't have to be internal," County said.

Self-provisioning banned

County said the institution had trialled allowing staff and researchers to provision their own infrastructure on-demand from the new, elastic data centres.

However, the institution back-pedalled after finding the feature simply led to more work for support stuff.

"Provisioning is still very techy in a way so it's hard for business people to interact to do that," he said.

Instead, all requests for new infrastructure from staff are channelled through the department, which can more easily map the request to the amount of available resource.

"By backtracking and getting an understanding of the real needs of the organisation area, understanding how long they need something for, and mapping that into one of our standardised set of images was much more valuable to us," he said.

"We spend a few days – two to three days, depending on the project – understanding what the need of the business is, then the actual provisioning [happens] at a click."