QLogic FabricCache puts poor SQL performance on the mend for Royal Bolton Hospital

An aging SQL cluster spread across slow physical systems was in desperate need of a performance increase to accelerate Royal Bolton Hospital’s largest SQL data warehousing operations. Royal Bolton’s two data centres are connected via a 1km fibre link and are synchronously mirrored for optimum resilience and failover capabilities. The SQL systems are critical to Royal Bolton, as they provide accurate and ongoing financial reporting used for calculating payments based on imported patient data. Storage subsystems were not able to fulfil the intensive data mining transaction requirements, resulting in excessive transaction latency coupled with lack of resilience, which led to gaps in availability.

“Financial planning and accurate patient analysis remain the lifeblood of any effective NHS trust. The accuracy, speed, and content of these resultant SQL reports were being hindered by low IOPS and high latency transaction times,” said Brett Walmsley, Royal Bolton’s CTO. “As it was, the two Dell servers housing SQL were already highly specified and few further gains could be made from additional disk subsystem enhancements. The resulting application response times could best be described as sluggish, and often unpredictable, while SQL shuffled across demand spikes and waited for the existing storage I/O to process the reports.”

A longtime advocate of QLogic infrastructure solutions, Royal Bolton was receptive to QLogic reseller partner NCE Computer Group’s suggestion to deploy QLogic FabricCache™ 10000 Series adapters. FabricCache is the industry’s first caching storage area network (SAN) adapter and features a design breakthrough that enables cache-based acceleration of virtualised and clustered applications without changes to existing server software or infrastructure.

“Royal Bolton operates in a highly virtualised environment. Data warehousing on the other hand, due to the vast amounts of I/O memory required, is not the best candidate for virtualisation, so instead we elected to bring it into our dual cluster,” Brett continued. “The SQL environment was therefore a perfect test bed to deliver FabricCache’s cluster-ready acceleration and resilience improvements direct to the users.”

Royal Bolton installed FabricCache into two Dell Rack R720 servers as a cluster-ready application accelerator. The unique design enabled FabricCache to simply appear to the SQL application and the OS as just another standard QLogic Fibre Channel adapter. Cache-based acceleration was targeted to specific areas where low latency operation would benefit the application most.

FabricCache application acceleration reduced SQL transaction times by more than 60 percent. Business intelligence reports that previously took 5.5 hours were reduced to a little more than two hours with no licensing or administration costs and no additional hardware or management overhead. In addition, the small outlay for FabricCache was fractional when compared to the expense of a broader storage infrastructure upgrade or the sharing limitations of standard, add-in flash products.

“We now have both FabricCache cards running on a two-node Window Server 2008 R2 cluster with a SQL 2008R2 clustered environment. It’s incredibly easy to use and manage,” Brett concluded. “We didn’t do anything apart from inserting the cards. No changes were made to native drivers or to the configuration. Once in situ, we simply specified the problem LUNS to target through a few simple clicks. The management GUI remains the same as all other QLogic adapters and is therefore incredibly familiar to us.”