When David Pickup first started investigating virtualisation technology, he had an inkling that it might help him constrain the sprawl of servers at East Lancashire Hospitals NHS Trust’s datacentre, which he oversees as the trust’s IT operations manager.

“We were running out of space and having to put in two racks each year, which was expensive on the power side and on air conditioning,” says Pickup.

The trust built a virtualised infrastructure in 2007 using VMware technology. Two servers run some 20 virtual machines, which support systems including BlackBerry Enterprise Server, departmental databases and the A&E management system for bed allocation.

But while virtualisation had helped to ease datacentre congestion, Pickup needed to be sure the system could meet the trust’s stringent data security rules.

“All data is confidential and we are not allowed to lose any patient information, so to the best of our ability we must achieve resilience,” says Pickup.

The trust created an iSCSI virtual storage area network (SAN), initially buying three 3.6TB nodes and one 9TB node from LeftHand Networks, later adding an extra 3.6TB node and a second 9TB node.

VMware Infrastructure 3 includes two business continuity technologies: High Availability, which automatically restarts virtual machines on another physical machine if a physical host fails; and VMotion, which allows the movement of a live running virtual machine to another host without downtime.

VMotion works with virtualised storage as most of the virtual machines are stored on the LeftHand SAN, so the two technologies are complementary.

However, virtualised storage appeared especially attractive when compared with the cost of HP’s Fibre Channel Enterprise Virtual Array (EVA). “The speed of EVA is an advantage, but it is cost prohibitive, with extra money necessary for training because our networking guys are not familiar with Fibre Channel,” says Pickup.

The trust is in the process of splitting its cluster between two locations, with each site having two 3.6TB nodes and a 9TB node asynchronously replicating data to each other to beef up its disaster recovery provisions.

The trust’s virtual server and SAN infrastructure means that if a virtual server dies, it can be rebuilt in half an hour. “I could lose a few drives and boxes and still have access to the data. If a physical server dies, you have to take delivery of a replacement, unpack it, install it in a rack, cable it up, all of which can take up to four days. Also, you have to recover the information which has hopefully been backed up on disk, which can take several hours,” says Pickup.