Keeping data storage healthy: University of New Mexico Health Sciences Center takes action to manage research data files efficiently.

All men may be created equal, but all data is not. True, all those microscopic magnetic fields residing on the drives of an enterprise's storage servers look pretty identical to the read/write heads, but the value that people assign to them varies greatly. Some of them are practically useless, whereas others contain vital operational data. Learning to treat these bits differently can result in considerable cost savings while improving users' speed of access.

This is where information lifecycle management (ILM) software comes into play. ILM software analyzes what is stored on the servers and allocates it to the appropriate storage device based on the organization's needs and policies. This includes, for example, how frequently a particular files is accessed, as well as requirements laid out in legislation such as the Health Insurance Portability and Accountability Act. For example, the data accessed most can reside on high-speed SCSI drives, while lower priority fries go to slower, but cheaper, ATA drives, or occasionally to optical or tape devices.

A Data Smorgasbord

The University of New Mexico's Health Sciences Center (HSC) in Albuquerque, N.M., is composed of a mix of educational, research and treatment facilities. Its three schools--the School of Medicine, which was ranked last year among the top 10 U.S. medical schools for training primary care physicians, the College of Nursing and the College of Pharmacy--enroll a total of 1,500 students, not counting those serving their residencies. HSC also has nine hospitals and clinics, including the 384-bed UNM Hospital, which was named one of the nation's 100 most-wired hospitals in 2002.

Providing IT services for 5,000 HSC students, staff and administrators comes under the duties of the Health Sciences Library and Informatics Center (HSLIC). It is a traditional library, but it also hosts standard computing services such as e-mail and Web services. HSLIC primarily uses Novell NetWare for its file and print servers and Microsoft Windows for its Web servers. All of its 8 TB of RAID 5 storage is locally attached, without using a storage area network (SAN) or a network-attached storage device. "We continue to find that local storage is very cost-effective, particularly when used in combination with secondary storage," says Barney Metzner, HSLIC manager of IT systems. "We've been able to amass a considerable amount of data without investing in a storage area network."

However, maintaining that structure requires a constant watch over all of the servers to make sure that they don't become full. HSLIC provides network storage to all users, including some who back up their hard drives to the network and many who generate huge quantities of research data: statistical files, PowerPoint files, medical images, numerous databases and other large files. Much of the research data has a relatively short life span, since they are constantly being replaced by new data, but other data files can go many months without being accessed, so they occupy space on the servers for a long time.

Volume Management

About five years ago, Metzner began examining hierarchical storage management (HSM) solutions to better manage HSC's growing storage needs. He recommended Managed Server HSM by CaminoSoft Corp., Westlake Village, Calif., and HSLIC bought the ILM software in April 2001. Managed Server HSM automatically migrates files from primary storage to secondary and tertiary storage based on policies set by the storage administrator. As with other HSM products, when migrating a file to secondary storage, Managed Server HSM replaces the file on the primary storage device with a "stub file," a small file that contains a pointer to the new location of the file on the secondary storage device. Stub files are completely invisible to users; if users access files that are in secondary storage, the files automatically migrate back to primary storage without users knowing it.

While CaminoSoft technicians installed the software, HSLIC's IT staff spent about one week establishing the three-tier structure. Metzner set unique policies for each volume based on what it held. In most cases, primary storage files that had not been modified in the last year would be sent to secondary storage. "Deciding on the rules for what you're going to migrate is a big part of the process," he says. "The limitations of the software such as the inability to migrate using complex rules combining creation date, last modified date and last accessed date mean that migrating and deleting just the files you want takes some careful planning."

HSLIC dropped the third tier of storage, however, because Managed Server HSM originally had limited support for optical and tape devices, and the supported devices did not provide the capacity that Metzner wanted for a third tier. Although later versions of the software addressed the support problem, they also contained a new feature that offered the ability to establish deletion policies. Rather than HSLIC moving files that had not been accessed for at least two years to tertiary storage, Metzner preferred encouraging users to save their files onto CDs or DVDs to keep the files available. Then, these files would be deleted from secondary storage (and their stub files removed from primary storage), although they would be available in different states of modification from HSLIC's routine backup tapes, if needed.

"A lot of the files are backups: out-of-date documents and research files, not core business data. Keeping them indefinitely is wasteful. It is easier to put a bit of responsibility on the users to properly manage important data that is rarely accessed. They should archive files off to optical media, tape or floppy if they need to keep them indefinitely but access them less than once every two years," Metzner says.

Staying Lean

Managed Server HSM manages nearly 4 TB of individual file storage. While most of the data migration occurs without administrator intervention, there still are cases when Metzner has to step in. Occasionally, for example, a primary storage server that holds up to approximately 2 million files will hit 80 percent to 90 percent utilization without containing any files that meet a condition for migration. In such a case, he will use the ILM software to bring the server's file volume down to 50 percent or 60 percent by manually migrating files to the secondary storage server, which accommodates more than 4 million files.

Metzner also used the software recently when moving files to a new primary storage device during a hardware upgrade. "It would have been painful to move all 250 GB," he says. "Instead, I used a more aggressive migration rule to drop the volume down to 80 GB of primary storage, so the conversion to the new hardware went much faster."

By continuing to keep primary storage cleaned up, HSLIC has managed its growing storage needs without adding expensive, high-performance drives. Instead, it just adds more near-line storage as necessary, which can be done without interrupting user access to active files. Now, when HSLIC upgrades primary storage devices, it does so to boost performance, rather than to add capacity.

"We won't run into storage problems for two reasons," Metzner says. "The first reason is our deletion policy. The second reason is that technology use is changing. More people are using enterprise applications that save data on large database servers, so they don't have to create a lot of separate spreadsheets or databases to share with co-workers. The files that people are creating tend to have a shorter shelf life and can be managed with our deletion policy. Secondary storage is growing, but our deletion policy will keep that growth manageable."

Metzner estimates that HSLIC spends about one-eighth as much for storage as similar institutions that have implemented a SAN with a capacity comparable to his center's file storage servers. "We have been able to offer our users nearly unlimited storage for their business, research and education-related files. We can do it without spending a lot of time and money adding additional capacity to our primary servers, employing more staff to train users about deleting files or restricting storage use. With the dramatic drop in the price of SAN solutions, we are looking at SANs as a way to increase performance, flexibility and reliability, but Managed Server HSM will continue to be a part of our strategy," he says.

For more information about Managed Server HSM from CaminoSoft, www.rsleads.com/404ht-200

SOURCE

Barney Metzner

Manager of IT Systems

Health Sciences Library and Informatics Center

University of New Mexico 

Albuquerque, N.M.

[e-mail]

PRODUCT/COMPANY

Managed Server HSM

CaminoSoft Corp.

Westlake Village, Calif.

www.caminosoft.com