SwiftStack object storage cures Oklahoma biomed's capacity, cost woes

Growing complexity of scientific data compelled the Oklahoma Medical Research Foundation to consolidate primary storage and archived backup on clusters of SwiftStack object storage.

The Oklahoma City-based nonprofit wanted to avoid vendor lock-in with scalable storage designed for commodity hardware. The plan was to retire EMC Isilon and Dell network-attached storage for a high-performance computing platform and phase out tape storage for internal server backups.

The Oklahoma Medical Research Foundation (OMRF) conducts grant-funded medical research on aging, cancer, cardiovascular disease and lupus. The institute's scientists are credited with more than 700 patents, including several federally approved therapeutic drugs.

OMRF CIO Brent Keck learned of SwiftStack's object storage software four years ago from colleagues at Fred Hutchinson Cancer Research Center, a sister institution in Seattle.

SwiftStack sells and supports a commercial version of open-source OpenStack Swift object storage, adding enterprise features such as file services and a management system.

After OMRF looked at SwiftStack and decided to implement it, the foundation's researchers and IT staff teamed to design the SwiftStack storage requirements.

"The real win is that SwiftStack allows us to solve our backup and long-term continuous data growth," Keck said. "We had administration and data scientists working collaboratively to solve a problem. That doesn't happen very often."

SwiftStack allows standard on-premises servers and hard disk drives to serve as a private cloud akin to Amazon Simple Storage Service. OMRF must retain certain data sets indefinitely for future reprocessing, such as data related to genomics research. The foundation previously used tape to archive and back up historical data associated with rare samples.

"Our tape drive was getting old and we needed to replace it," Keck said. "We had a choice: Spend $60,000 to $80,000 on a new tape system, or use that money to get off tape for long-term storage of data that doesn't need high performance."

OMRF replaced EMC Isilon with high-density Supermicro servers at its three regional data centers. It has nine SwiftStack nodes equipped with nearly 280 3 TB and 6 TB SATA drives.

Each SwiftStack object storage node is configured as a node and a proxy. To ensure disaster recovery, each three-node cluster replicates data between the data centers via 10 Gigabit Ethernet.

The SwiftStack deployment supplies more than 1 PB of raw storage, which translates to about 350 TB of useable capacity. Roughly 200 TB of storage is consumed by scientific data and 150 TB of space is used by a Commvault backup system.

OMRF depends on grants and donor funding, and Keck said that makes it difficult to devise a multiyear storage budget. Using SwiftStack object storage allows OMRF to elastically scale capacity by adding enterprise drives to the Supermicro chassis.

"The nice thing about SwiftStack is that it's pay-as-you-go," he said. "The upfront cost of building these arrays, even with commodity hardware and drives, was equivalent to buying a 7 PB array. We didn't save money on the initial buy. Our payoff will come in a year or so when we can avoid the cost of having to buy another array."