Bioinformatics specialist Oklahoma Medical Research Foundation (OKMRF) brought in Isilon Systems Inc.'s clustered network-attached storage (NAS) system this year to support a new gene-sequencing tool, but found the clustered NAS system could also support a new VMware Inc. vSphere 4 deployment with adequate performance for both environments.
OKMRF software engineer Stuart Glenn said the gene sequencer and Isilon storage to support it were funded in part by a grant through the American Recovery and Reinvestment Act of 2009 (also referred to as the stimulus package) put into effect earlier this year. Previously, the foundation used a hodgepodge of home-built and commodity NAS servers, direct-attached storage (DAS) and Apple's Xserve RAID SAN storage.
Glenn said because a typical gene-sequencing run can take up to a week and produce between 2 TB and 6 TB of raw images, it's important to be able to expand storage capacity quickly. The foundation was also planning to roll out vSphere 4 server virtualization, and sought to kill two birds with one stone.
"I kept coming back to Isilon because of its throughput and clustered nature," he said. Glenn didn't have any specific performance benchmarks handy, but said "We're pushing our Gigabit Ethernet link."
BlueArc could push that link as well, but Glenn said Isilon's pricing was more attractive. He was also interested in the choice of attachment methods for hosts to the Isilon cluster, which would allow the gene sequencer and vSphere to share the storage.
"With their SmartConnect option, you can do a virtual connection so you see one DNS name, or you can have [hosts connect to the Isilon system] round-robin, or you can have connection and failover set up on particular nodes," he said.
He chose the latter option to keep the VMware and sequencer from competing for resources. "We can have the sequencer on one node, five to 10 Windows clients and Mac as well, plus Linux compute nodes, all able to connect to a different node of the cluster," Glenn said.
While Isilon separates front-end connections, it parallelizes data and cache among the system's nodes, which are interconnected with InfiniBand to speed aggregate performance for all of the foundation's hosts. Having a single system to manage has also proven much more efficient, Glenn said. "One of the biggest things for me, since I'm kind of the only person who manages all this, is that I don't have time to sit all day provisioning LUNs [logical unit numbers], and configuring Fibre Channel volumes."
OKMRF is also taking advantage of vSphere's new native support for NFS data stores. "It doesn't perform as well as Fibre Channel SANs, but…the ease of management far outweighs that," Glenn said.
Glenn would like to see Isilon take a page from the Sun Zettabyte File System (ZFS) and add support for native compression or data deduplication. Although Isilon partners with primary storage data deduplication startup Ocarina Inc., other Isilon customers, including Zoic Studios, also place integrated primary storage data reduction on their wish lists.