PACS networks offer important advantages over film-based systems, but implementing them presents a number of tough questions. For example, which storage medium should be used to archive the boatloads of image data -- perhaps 20 GB a day, or 4 terabytes a year -- that come with running a large-scale PACS network? Storage capacity, speed, cost, and space requirements are important factors, as is the need for flexibility.
Massachusetts General Hospital PACS administrator Patricia Whelan believes that the choice of storage media is one of the most important PACS-related decisions an institution can make. MGH, which treats nearly a million patients a year, will eventually need 40-60 terabytes of storage capacity to meet state statutes for retention of medical records, she said.
"It's important to have a strategic plan, and to make sure all of your moves fall under the plan," Whelan said.
Developing a strategy means determining how much storage will be needed, how much it will cost, and which tradeoffs in price, performance and reliability will work best for the facility, Whelan said.
CD-ROM is an option that has been used in PACS archiving, but the disks hold only 600 MB of data -- or the equivalent of about 12 chest x-rays. Combined with a slow data transfer rate, low storage capacity makes CD-ROM a poor choice for PACS archiving, Whelan said.
DVD, the format you might rent your movies on, is more flexible. The disks hold three times the data of CD-ROM disks -- up to 1.7 GB each -- but their data transfer speeds are rather slow with current technology, Whelan said. Technology will soon speed up transfer times, making the format more practical in future installations, she said. There are also standardization issues that need to be resolved before DVD can become a serious contender in PACS storage.
Two other formats do present a viable storage choice for large-scale PACS archiving. These include MOD (magneto-optical disks) and DLT (digital linear tape), both of which are in current use at MGH, Whelan said.
The team compared the performance of both formats, and found that each has its strengths and weaknesses. The comparison was based on five criteria: performance, cost, capacity, reliability/scalability and redundancy (see illustration I).
MOD had a slight edge in performance. However, she said, "People tend to focus on the retrieval or read function; they don't focus on the write function. MOD retrieves twice as fast as DLT, but DLT writes twice as fast as MOD." In terms of capacity, DLT was the clear winner, with 35 GB storage capacity per tape compared to MOD's 2.3 or 5.2 GB.
"Both drives are prone to disk defects and drive failures. However, MOD was more consistently reliable, with fewer drive difficulties and fewer media defects," she said.
Scalability relates to the institution's ability to continue to archive. MOD's jukeboxes require far more space to archive than DLT, making DLT the clear winner in this area. On the basis of space requirements and cost per bit, MGH anticipates that it will phase out its MOD storage to embrace DLT technology and take advantage of the latter's rapidly dropping prices.
The two media finished about even in redundancy; that is, the ease and efficiency of creating backup files
RAID (redundant arrays of inexpensive disks) is a crucial component of any archiving strategy, Whelan said. Files are archived in RAID for a period of time after they are created in order to limit the number of file requests from the image archive. In addition, RAID allows the user to split or copy data on several drives, avoiding data loss even if a disk fails.
How much an institution decides to invest in RAID storage directly affects its archive capacity requirements. Because the cost for various RAID storage options are dropping rapidly, the size of the typical RAID configuration is growing, she said. Some institutions are now implementing RAID capacity of 2-3 terabytes in order to handle as much as 80% of their file requests online, eliminating the need to access archived data most of the time. RAID is faster, but more expensive per MB than other archive options (see illustration II).
Within a few years, increases in Internet bandwidth capacity are expected to make off-site storage through ASPs (application service providers) practical, Whelan said. Besides eliminating the need for storage space, infrastructure, and administration costs, she said, this option eliminates the need for large storage-related capital expenditures.
With off-site storage, "you don't have to come up with half a million dollars for an archive. You can have $2 a study fee-per-use, and pay for it out of your operating budget," Whelan said. "The reality is that someday we'll all be doing it, but for now the economic incentive is for small and mid-sized hospitals to participate. The larger organizations won't be as quick to go to an ASP model because they've already bought their on-site archives."
In addition, holographic glass-disk data storage systems are being developed that can store at least 10 times the data of other media in the same amount of disk space, Whelan said.
A leader in the quest to commercialize holographic storage is Aprilis of Cambridge, MA. The company is fine-tuning a storage scheme based on cationic ring opening polymerization technology developed by the National Storage Industry Consortium (NSIC). Specifications for the work-in-progress Aprilis system call for capacity of one terabits (1 trillion bits) per disk, random access, and a sustained data rate of six gigabits per second using a single optical head.
"Because of the incredible density of these things, it's going to shrink the amount of size and space you need for your archive. The difficult thing about all storage changes is going from one media to the next. How do you migrate the data from a disk onto the holographic storage? These challenges will continue to exist," Whelan said (see illustration III).
While holographic schemes edge closer to the market, magnetic storage technology is also advancing rapidly. Last year, for example, Sony succeeded in putting a million bits of information on a square inch of tape, using magneto-resistive heads in a helical scan tape mechanism. A few PACS vendors offer the firm's Advanced Intelligent Tape (AIT) as a storage option. In addition, TeraStor's near field recording (NFR) technology offers a disk-based method of increasing areal density capacity. Expectations are that, like holographic systems, magnetic devices will also reach 1-terabit capacity levels before the decade ends, although data transfer speeds will likely remain slower than holographic.
In conclusion, Whelan said that storage strategies vary depending on an institution's threshold for image retrieval cost and expansion needs. In order to make good storage decisions, an organization must know what it wants, communicate its vision clearly to vendors, and remain flexible as new technologies are developed and prices for older technologies fall, Whelan said.
"What do you want for your organization?" she asked. "Seven times cheaper or two times slower? What concessions will you make? Storage is not a one-time decision," Whelan said. "It must be updated and changed as your organization grows."
By Eric BarnesAuntMinnie.com staff writer
September 19, 2000
Click here to view all active PACS Digital Community discussions
Copyright © 2000 AuntMinnie.com
