An international law practice swapped four racks of servers for a boxful of blades running virtual servers. The...
results of blade virtualisation are improved billing throughput, greater service reliability, and huge savings in ongoing maintenance and capital expenditure costs.
Rouse, which specialises in international intellectual property, reckons that, as a result of its move, it's achieved a 75% reduction in the time required to perform key business processes, such as accounting and billing.Spotty system service
Rouse employs some 550 people worldwide in 20 offices, but its IT function is located in London's Docklands. According to Jonathan Bruce, the global head of IT, as staff numbers grew, it put more pressure on IT to the point where systems became unresponsive to users.
"We built a data centre in 2006 with a 42U rack of servers, but they were on the limit of the UPS [uninterruptible power supply]," Bruce said. "We have space constraints, and we were aware of power and cooling constraints too. And there wasn't any power management in the older servers."
The combination meant that Rouse spent a lot of money on an increasingly unreliable infrastructure.
"It started to fail over frequently -- almost weekly," said Bruce. "It was a lot of work for IT to come in overnight and reboot. And it started to affect access, especially since we have access requirements across the globe.
"The unreliability of the system was primarily caused by an Adaptec Snap Appliance 18000 not having the I/O capacity -- read/write failures would sometimes halt the system totally -- when it came to the month-end billing cycle. The previous HP server hardware was more than 5 years old and was supporting 40% more fee earners in 2008 than when provisioned in 2004.
"That was the pain point," said Bruce.
Getting started with blade virtualisation
So Rouse decided to upgrade the system to blade servers hosting a virtual infrastructure, all with built-in power management. "That was where virtualisation came in, with a blade infrastructure that we knew would be powerful enough to run our practice management system," said Bruce.
Previously an HP house, Rouse jumped the tracks and went to Dell to lease its hardware. "The decision came down to price," Bruce said. He described the Capex savings as "eye opening."
With the move came a decrease from 70 physical servers to 20 as well as a decline in power draw: Bruce said that the same UPS -- an APC Symmetra -- that previously had enough capacity to support the system for a few minutes will now run "for hours."
The company had used VMware for about three years, so the move included an upgrade to ESX Server on most of the blades, all under the management of VMware vSphere. "We're looking towards a full-on automated system in future," said Bruce.Migration
"We had virtualisation and storage area networking skills in-house, so we did 70% of the work, while the new vendor, Dell, did the rest," said Bruce. "We did a lot of P2V [physical to virtual] and gave the old servers back to HP when the lease was up. If they weren't on a lease, we disposed of them to a Waste Electrical and Electronic Equipment (WEEE) disposal agency."
While the low points of the move were the long hours spent at weekends setting up and configuring the new systems, Bruce said that he didn't anticipate having to work too many weekends in future.Business benefits of blade virtualisation
And most important, the business benefits were quantifiable: "At Rouse, the first two weeks of the month are when we send our bills out," said Bruce. "We were generating about four to six bills a minute. Now we're doing 18 a minute. Checking takes less time, as does debt ageing; this alone was taking about 14 hours and it now takes under two." Overall, the time taken to perform this kind of process decreased from nine-plus hours to less than two.
Additionally, said Bruce: "Aged debt and work in progress calculations were run as an overnight process and took 12-14 hours, affecting backup schedules and China's start of day, which meant it was very slow for the end users. These processes now take under three hours and can be run during office hours."
Service-level agreement costs dropped as well, Bruce said. At £40,000 annually, Rouse now pays less for hardware leases and maintenance contracts than it paid before just for hardware leasing.Disaster recovery
In parallel with the server upgrade was a new disaster recovery project, the aim of which was to replicate the system off-site at the company's disaster recovery colo facility in Wapping. The company bought a new EMC SAN and Mirrorview replication software as well as backup software from DoubleTake.
Rouse deployed VMware Site Recovery Manager to manage the failover process, which takes under an hour. It's well within the organisation's recovery time objective of 12 hours, said Bruce, and quicker than the recovery times achievable through its previous, tape-based DR strategy, which sometimes took several days.
"Previously we used tape and couldn't be certain in the event of a disaster that we'd get everything back up and running," said Bruce. "VMware allows us to simulate fire drills and bring up the replicated site in a virtual bubble and test the validity of the data."
As a case study in possible savings when swapping from physical to virtual servers, Rouse is a good demonstration - although much of the energy savings also result from the deployment of more modern hardware, which delivers more performance per watt.
Manek Dubash is a contributor to SearchVirtualDataCentre.co.uk.