The trust comprises three major hospitals in Worcestershire and some 90 remote clinics and surgeries with 4,500 staff serving a population of more than 500,000 people. When the trust was formed by merger of existing NHS organisations in 2003, it inherited an estate of 200 servers each with direct-attached storage.
The trust's IT team initially opted to run FalconSstor IPStor software on two servers, one at Worcester and one at Redditch in front of two Nexsan SATABeast storage arrays with 3TB of capacity at each site.
"We didn't migrate all servers at first because many were tied into external support arrangements," said Peter Lowe, centralised IT services manager with the trust. "But as they became free, we moved the data over."
As a means of reducing costs and avoiding the management workload and training that would have resulted from linking its numerous sites using Fibre Channel, Lowe's team chose to use iSCSI connections from the remote sites to the FalconStor devices with Fibre Channel from there to the Nexsan arrays.
Increasingly popular, iSCSI SANs use existing Ethernet cabling and networking skills and carry SCSI commands within an iSCSI wrapper. iSCSI is not generally recommended for high-transaction data centre use cases, but for relatively low payloads over longer distances -- especially for connections to branch office environments – it can sidestep prohibitive Fibre Channel infrastructure and training costs.
When the Worcestershire project began, iSCSI was not fully accepted and doubts were raised about its suitability for distance transport of data, said Lowe. "When we started, iSCSI was frowned upon," he said.
"Some providers were less than enthusiastic, saying its performance was poor. I didn't think it was a gamble at the time but many thought it was, so it's nice that it has paid off. And there turned out to be no performance hit with iSCSI – it has worked a treat and there would have been no benefit to go with Fibre Channel, especially as iSCSI will soon be faster with 10 Gigabit Ethernet."
FalconStor's IPStor is a storage management product that allows users to link heterogenous storage arrays and incorporates features such as replication, mirroring, snapshotting and continuous data protection, which the trust has now implemented. The trust has also since upgraded to a two-node IPStor cluster at each site to enable failover.
Lowe decided against using only one vendor for its storage management and arrays so he could better take advantage of developments in the marketplace. "We looked at EMC/Dell but we didn't want to be tied to one vendor," he said. "I prefer to keep an eye on what's going on with the market and knew that as the technology matured we'd have chance to exploit falls in price."
He opted for Nexsan arrays on grounds of storage density and now has 200 TB of capacity on that vendor's SATABeast products.
Is there anything FalconStor could do better? Lowe would like more fine-grainedness of control over data defragmentation on drives. "I would like to see LUN defragmenting, for which EMC has the capability," he said. "I like to keep things optimal to limit performance degradation -- after many changes on a drive things can get messy. The only way to do it with IPStor is to copy a LUN to a different drive, in the process of which the data is defragmented."