Data backup and archiving often play distinct and complementary roles in data protection, but for the University of Michigan Health System (UMHS), a tiered data archiving system has almost completely replaced the role of backup.
UMHS uses a mix of disk and tape tiers to store more than half a million radiology and cardiology exams. The tiers include an IBM Corp. DS4800 Fibre Channel storage area network (SAN), two Oracle StorageTek SL8500 tape libraries and 12 Picture Archiving Communications Systems (PACS) tied into a grid storing patient records in large files.
UMHS includes hospitals, health centers and clinics, the University of Michigan Medical School and its Faculty Group Practice, clinical activities of the University of Michigan School of Nursing, and the Michigan Health Corp.
Steve Ramsey, director of image management and computing services for the radiology department at the university, said his department is responsible for the diagnostic imaging storage for the health system. He needs those images stored long-term and available for quick retrieval when a doctor needs them.
"We're mostly interested in archiving," Ramsey said. "All we're backing up is database applications, and those are small. When it comes to medical images, we're just looking at archiving. We keep two copies, but I don't consider that backup.
"We use disk to retrieve data that's been archived within 12 to 18 months and tape for anything older than that. We archive everything forever. We don't purge any medical imaging data."
Ramsey said the health system adds about 30 TB of data compressed at 2.5-1 ratio per year and is growing at about 20% a year. He said tape helps him keep up with that growth at a relatively low cost.
"Our biggest challenge is handling the growth of data and doing it in a way that is economically feasible for us so we don 't have to worry about running out of storage space, and not have it be a financial burden year after year," he said. "We're always having to upgrade and add capacity. What I like about tape is it has the capacity to grow without requiring large costs from year to year, and I don't have to worry about finding extra space in data centers."
But while tape lets UMHS store and expand its capacity easily and at a relative low cost, the quick recovery capability of disk plays a crucial role in the process. Ramsey said he uses the IBM DS4800 Fibre Channel SAN "for very fast retrieval of the most recent data."
Ramsey said UMHS automatically moves data based on its age and the last time it was viewed.
UMHS keeps copies of its data in two data centers about five miles apart in the Ann Arbor, Mich. area. It replicates between data centers, which also provides disaster recovery. "We have redundant systems, we can switch over to another site if we need to," Ramsey said. "But I also want two copies of all data."
A key aspect of the UMHS data protection system is the ability to quickly produce records stored in its vast archive.
"Images are accessed 24/7, and sometimes they have to be restored quickly and there's no real pattern to it," Ramsey said. "A patient may come in through emergency that hasn't been seen in five years. A physician needs to look at a five-year-old image for comparison. That's why we need to keep our data forever, we don't know when the data will be needed. And it's not just defined to one department, it can come from anywhere throughout the health system."