30 servers to 7: BUPA redoes virtualisation
Most IT teams spend 90 per cent of today making sure that tomorrow is not worse than yesterday, according to Paul Berryman, BUPA Aged Care CIO.
Any project that will change the 90/10 balance is worth doing, in Berryman's book; he only has a team of six people to service the 4800 staff and 4200 residents at the company's 48 aged care homes.
The company has a number of projects on the boil, including revamping information and security policy to be in line with global requirements, upgrading its enterprise resource planning (ERP) system and standardising 55 telephone systems into one. But the one project that is the foundation for all of these is its server-virtualisation project.
Previously, the aged care homes had been running virtualisation — but had been doing it badly. Although servers were virtualised, they each had their own physical servers; so there were 30 hosts all running one guest.
This way of doing things had a poor return on investment, made monitoring and detection of faults difficult and made the environment in general too complicated for the team, so that rebuilding a server took hours.
He wanted to have infrastructure that was expandable, that saw optimised use of hardware and enabled business continuity and disaster recovery. The latter was particularly important, because of power outages and floods. If systems went down, not only was it not ideal for taking care of patients, but it also stymied relatives' attempts to find out what was happening.
Berryman had two options. The first was moving to cloud-based services. However, his opinion of the cloud is that it is either "thin and wispy and not really there" or "dark and gloomy".
His other option was to approach one of multiple vendors and ask about virtualisation. He was concerned about the fact that Microsoft was a relative newcomer to the market, and thought that desktop virtualisation was Citrix's strength, which is not what he wanted. VCPro, on the other hand, a VMware specialist, attracted Berryman, because he, coming from a specialist organisation, believed that specialists would do what they did well, and not try to sell extraneous other products and services to him. He also liked the fact that VCPro was happy to work with other vendors, for example Dell, on the project. So he decided to conduct the implementation with VCPro's help.
Berryman has moved almost everything onto virtualised servers now, including the company's ERP Solution, payroll and HR systems, time and attendance, email, file and print, clinical systems, internet, an SQL database cluster and test-development and training instances.
There are seven virtual server hosts running over 50 guests, with automatic failover when a physical hardware unit goes offline, automated load shifting and quick deployment of servers — the admins can build a new server in about half an hour.
The reasoning for Berryman's virtualisation project was to improve services and free up his staff to do other things, such as the ERP upgrade, and implementing desktop virtualisation, for which he has not yet chosen a vendor. It was with these benefits in mind that he took the project to management for funding. He admits now that perhaps this wasn't the best avenue he could have taken. He believes that he would have received approval much faster if he'd talked up the dollar savings. As it was, it took much longer for him to get through the due diligence than it did to actually implement the project, which he said took around a month.
The implementation cost $100,000, and he expects to get the return on his investment in two and a half years, as he's saving $40,000 per year by not having to replace the rest of the 30 old servers that were so expensive to maintain.
But it's been about much more than savings; it's been about improving services, according to Berryman.
"I think my single point of failure is now my staff," he said. "Their documentation is now in their head."
Even so, when one key member of his team went on holiday, and there was a problem where the managing director suddenly wasn't getting any emails, the fix went without a hitch (or would have, if the employee had remembered to pass on the IP address for vCenter). Once the team had that address, it was able to remotely restart the server that had been acting up.
"Had we known the address, we could have done it in minutes, and it could have been done from home," he said.