Storage virtualisation delivered by FalconStor's Network Storage Server (NSS) product has helped Northern Rail...
cut disaster recovery restore times from three days to two hours by replicating data between the company's two key data centres.
The 12-month project -- which saw NetApp and Hewlett-Packard (HP) products rejected -- was instigated when Northern Rail's board decided the company needed to drastically cut recovery time objectives (RTOs) from between 48 hours and 72 hours to less than 12 hours.
Northern Rail employs 4,800 people and provides rail services to a population of approximately 15 million people. Protecting SAP running on Oracle databases was the key aim of the disaster recovery project and with existing RTOs it was estimated that a lot of data would have to be re-keyed into its systems.
"Disaster recovery took too long," said Edwin Lloyd-Jones, server manager at Northern Rail. "We were backing up solely to tape, and if disaster struck we'd have to wait for hardware to be sourced and data to be restored from tape. That could take anything up to three days and the board had decided we needed to get that down to less than 12 hours."
Specific disaster recovery concerns included Northern Rail's main data centre at York being in a regularly flooding area and also being sited at a key railway junction that was potentially subject to a security cordon in case of any threat to the rail infrastructure.
Prior to the project, the company was solely using a traditional backup regime as data protection. It used Symantec's NetBackup to copy data to tape from an HP EVA6000 supporting 45 servers at the York office and from direct-attached storage (DAS) at its Manchester site.
Following an evaluation of three vendor products, Northern Rail implemented two FalconStor NSS storage virtualisation devices at York and one at Manchester to replicate SAP data between sites. Storage at the Manchester site was also upgraded with the implementation of a 5 TB Dot Hill 2730 SAN to replace the existing DAS setup.
The FalconStor devices are storage product-agnostic and can replicate data between the two vendors' SANs. "We're replicating delta changes of SAP data and VMware images continuously. Also time-marked changes, which is what we'd roll back to in case of a disaster recovery invocation," Lloyd-Jones said.
Lloyd-Jones' team also evaluated the use of NetApp devices to effect replication between the two sites, but these were rejected on cost grounds. It also considered adding another HP SAN in Manchester.
"We looked at getting another EVA at Manchester and replicating between them, but we wanted to be able to be vendor-independent. Also, it was a very costly solution to our situation -- good, but expensive," Lloyd-Jones said.
"FalconStor is storage-independent, which was important for us as we want to retain the ability in future to choose storage that suits our needs and not be restricted to one vendor," he added.
Benefits of the storage virtualization implementation are that RTO is now down to approximately two hours -- a much shorter disaster recovery restore time than even Northern Rail's board set as a target. Lloyd-Jones is also happy that FalconStor has proved it can scale to the addition of further SANs from different vendors.
His only gripe is that FalconStor currently only supplies agents to use with SAP's native database, MaxDB. "Full SAP integration would be nice," Lloyd-Jones said. "Currently, NSS doesn't have agents for SAP/Oracle. So, what we do is to snapshot Oracle directly, which we've thoroughly tested, and it works, but it's not SAP certified."