For Insurer Admiral, installing more physical servers would have made its datacentre unmanageable, and the cooling systems would have been unable to cope. So the company turned a local integrator who persuaded it to run with a VMware virtual infrastructure.
The group consists of some of well-known insurance brands, including Admiral, elephant.co.uk, Diamond, and Bell. It employs over 2,000 people in its offices in Cardiff and Swansea, and currently insures around 1.3 million motorists.
Admiral’s datacentres, based in Cardiff and Swansea, required a major overhaul. Hosting both internal applications as well as customer-facing systems, it was coming under increasing pressure to meet the demands of the business and the customers that it supports. The nature and growth of Admiral’s business demanded frequent additions to the Intel server infrastructure, and the rate at which the server estate was growing was becoming unmanageable.
In addition, the existing air conditioning system was no longer able to cope with the power and cooling requirements that these new servers would have.
According to network manager Paul Connah, the main issue was one of server sprawl - the main datacentre in Cardiff was bursting at the seams.
“In order to support the business, new applications were being deployed on individual servers, but this was causing a problem in terms of power - the building that we were in could not supply any more electricity to us,” said Connah. “The other problem was time: commissioning a new server would take about three weeks in total, from provisioning to final deployment, which was far too long a process to deliver the requested service.”
If things were to change, Connah needed to sell any changes internally before doing anything. He approached it researching the market and evaluating alternative approaches to hosting applications.
“As a solution to our problems, virtualisation was a no-brainer. By consolidating our servers, we could reduce the amount of power and space that our IT infrastructure would take up. The speed of rolling out new virtual machines was a massive plus as well - we could have a new server up and running from a template in minutes.
“As for selling the project internally, it would have been more problematic if we had not undertaken this project. The rate at which our business is growing and the IT requirements to support this growth meant that we required a new infrastructure strategy. We looked at the costs of implementation, and these were easy to justify over a three year timescale."
Connah and his team decided to use blade servers as the hardware basis for the project, largely for space reasons.
“We have three IBM BladeCenters, attached to two DS4500 storage arrays. On the virtualisation front, we are currently running VMware ESX Server 2.5.3; we are planning to roll-out VI3 in the near future.
“For the initial pilot, the hardware was commissioned and built along with our infrastructure partner Apex. We tested the configuration they had put together to make sure that it suited us, and then moved into our main site. As part of the roll-out, we shifted applications from physical to virtual (P2V) machines and did some out-of-hours testing to see that it was getting the right levels of performance. We would then roll back to the physical environment until we were happy with the performance of that application.
“Once everything had been tested, the application would be moved to a production virtual machine. We used PlateSpin PowerConvert for the P2V conversions,” said Connah.
Connah said that the selection process was surprisingly fast, at about six to nine months, including an evaluation of the hardware form factor and the suppliers.
“The main criteria for us were whether to virtualise our systems or not, and the management tools that were available once we decided to go down the virtualisation route. We also had to make the right choice when it came to the partner that we were working with; project management skills were a critical factor when it came down to it, and Apex were able to demonstrate this to us.”
Admiral chose IBM Director as a management tool for the new system. It was “very impressive”, according to Connah. “We can have an overview of both our physical and virtual server estates. The Intel-based blades and storage arrays fit in with our existing System I servers as well, and we have a complete overview of everything that is running in the datacentre.”
The server migration process was “steady,” said Connah.
“Using PlateSpin, we’d go through several P2V migrations until we were happy with everything. We could test the application thoroughly within the virtual machine, and make sure the switch-over to the production environment would go smoothly.”
Problems were few, and Connah attributed this to the detailed planning that took place before implementation.
“As this project was mainly owned by IT, it all went quite straight-forwardly. The main area of discussion was the initial project plan, and a lot of work went into this so that everyone was aware of the process and happy with it. Having a plan in place like this is the best advice that I could give for anyone else looking at virtualisation,” he says.
“The testing that we carried out ensured that we were careful on how we managed the implementation, and that no errors could creep in during the project.”
While Connah couldn’t provide details of the exact savings resulting form the migration to a virtualised infrastructure, he identified where the savings were made.
“Our maintenance overheads have been reduced, because we only have one set of hardware to support. We can also manage the physical and virtual systems from a central point. Secondly, the amount of space and power that we are using is lower - the new servers that we have are working that much harder, but because we only have a few of them the overall amount of energy used is much lower. Thirdly, we can react much faster to the needs of the business - as an organisation, we can be more agile in our approach and we are not constrained by the time taken to procure assets.
“This project has addressed many of the issues that we had as an IT department,” said Connah. “Power and space requirements have dropped considerably, and our administration overheads have also been addressed. The focus for us going forward is how we can meet strategic needs and help to grow the business, rather than reactively fixing problems. Our virtual IT infrastructure means we are well equipped for the foreseeable future.”