According to research conducted by the Enterprise Strategy Group, over 50% of all corporate data involves "fixed...
content," such as documents, images, e-mail and other data. These files are only accessed infrequently but they must be readily available when needed and retained for prolonged periods of time. Corporations are also experiencing up to 90% annual growth in fixed content volume. Such enormous growth puts increased pressure on storage resources, making content-addressed storage (CAS) technology particularly attractive to busy storage administrators. CAS is already being used to address specific storage needs.
Saving time and space
When RadioFrame Networks Inc. sought to expand their IT infrastructure, they faced the growing challenge of backup and restoration. The issue was more than just protecting about 1.5 terabytes (TB) of corporate and development data. They needed a means to restore quickly, and locate older file versions without long and arduous tape examinations. "We needed the ability to restore information in seconds," says Jason Hardy, network analyst, level 3 at Radio Frame Networks. "The ability to actually roll back to older versions of the information was key."
After evaluating a variety of tape systems and other CAS storage products, the decision was made to implement the Axion system from Avamar Technologies Inc. because of its speed and data reduction capabilities -- more than 1.5 TB of data is now being backed up to only about 800 GB of space. As a private company, RadioFrame Networks did not see regulatory compliance as a major driver for CAS technology, though Hardy recognizes the added value of such capability.
Hardy appreciates the fast backup/restoration and version control capabilities brought by the Axion, though he notes that data management tasks can be a bit cumbersome. "It's very bulky," he says. "It's having to hop in and out to several different applications to get the information you're looking for." It's a problem that seems to permeate the industry. "I don't think that the vendors out there have done a very good job of bringing it all together into one application." Hardy is expecting better management tools for the Axion to appear sometime in 2006.
Hardy is also hoping for more speed (especially across remote links) noting that current replication speeds are limited to 10 GBph -- primarily due to file system limitations. "It's just taking a very long time to replicate all of that information we have because of just the file system bottleneck at that level," he says.
In addition, Hardy feels that CAS support for Exchange needs some improvement. "It's a little bulky when you're trying to do message-level backups because it actually opens up every e-mail and every mailbox, and your Exchange system takes a hit on that [during backup]," he says, though improvements are expected in Avamar's version 3.5 software release.
Hardy advises readers to look far ahead when considering a CAS product. "Really pay attention to what your needs are now, but also try to forecast what your needs are going to be years from now," he says. "The last thing you want to do is be switching out technology and hardware two years from now because what you're putting it in isn't really cutting it anymore."
Meeting regulatory requirements
In the banking world, check handling has demanded significant attention. In the past, check images were relegated to microfiche and stored in a vault. Later on, the images were recorded to DVD. But recalling check images proved to be a labor-intensive process. Requests had to be made, the media had to be pulled and the correct image could then be located. Banks realized that check images could be digitized and moved to long-term, readily accessible disk-based archival storage systems, such as EMC's Centera. CAS helps the bank to meet regulatory retention needs and provides a far better level of service.
"We were able to replace the need for DVDs with Centera," says Paul Stonchus, first vice president, data center manager at MidAmerica Bank. "And now for all intents and purposes, the end user is totally unaware that the image they're receiving is a year old, but the response time is less than one half of a second."
Stonchus underscores the spiraling need for archival storage, noting that his utilization of CAS space is escalating at a significant rate. MidAmerica's first Centera was installed in March of 2004. Since initial installation, three additional storage servers have been implemented -- bringing the total Centera space to about 4 TB today. Over the next 12 months, Stonchus expects his CAS capacity to double again. Plans are also underway to establish another Centera platform at a remote location for off-site data replication.
While Stonchus expressed concern over Centera's up-front acquisition costs, he's quick to note that the total cost of ownership has been comparable to tape-based systems. Storage can be added cost-effectively and administration is very light. "The number of hours per week spent administrating Centera is maybe one half hour -- tops." Even firmware upgrades have been nondisruptive.
The biggest challenge that Stonchus sees with CAS platforms is getting vendors to support the storage systems with their software applications. "Talk to application developers and find out if they're prepared to write/read data directly from a Centera [or other CAS] device," he says. "Otherwise you end up having to put in a gateway solution."
Significant cost savings
Healthcare facilities deal with high volumes of fixed content such as patient records, X-ray images and MRI data. Placing this fixed content on a CAS infrastructure offers important cost and service benefits to organizations like Banner Health. "WORM [write one, read many] is exactly what radiology applications are all about," says Tim McFarlan, systems director, technology management at Banner Health. The trick for McFarlan was to find a CAS platform that proved suitable for Banner's Picture Archive Communication System (PACS) application used to manage and share digital radiology images across Banner's 21 hospitals in seven southwestern states. Today, McFarlan uses Bycast StorageGRID to handle about 175 TB of PACS data, covering about 10 hospitals. He expects to complete the digital conversion across all of Banner's facilities in the next two years.
The case for CAS technology for radiology images was a strong one. McFarlan explains that conventional film-based X-rays presented many drawbacks. Rising film costs, slow and messy development processes, environmental concerns over chemical disposal, and the inefficiencies of film transport and long-term storage all played heavily into the move to CAS. "Eliminating the film costs and the film-related costs, and the productivity gains in terms of the throughput (as far as how much faster we could process patients) came back as a cost reducer," McFarlan says, noting that digital storage also helps Banner to meet its legal and compliance obligations.
But Banner's road to CAS has been turbulent. Early CAS implementation choices with another turnkey vendor solution resulted in connectivity and other IT issues, as well as serious proprietary limitations that impaired interoperability with other PACS vendors. After reevaluating its needs and requirements, Bycast StorageGRID was adopted for its versatility and scalability.
"We could only go by aging of studies as a criteria for off-loading images," McFarlan says. "Bycast had the neat feature that we could use the metadata which is embedded with each study where we could treat different types of studies differently." Bycast's grid storage approach also allows for fault tolerance, allowing copies of data to reside in multiple locations -- hospitals can continue to operate even if Banner's MAN is interrupted or trouble strikes one storage node.
Although the move to CAS demanded careful, transparent data migration, McFarlan reports that the technology is performing well and the organization is generally satisfied with the transition thus far. "We did come through a very rocky road," he says. "We've had to ramp up rapidly and we've made some mistakes but overall I'm really happy that we've settled on this technology." His major hope for the future is to see added flexibility with metadata and archive change management to allow for more convenient data updates (e.g., patient name changes) as the need arises.
Go to the next part of this article: Content-addressed storage: Future directions
Or skip to the section of interest: