EMC's Centera makes play for 'fixed content' with lower cost storage

Amicas boasts first Centera-based PACS

The proliferation of digital data is driving the need for storage--on-site, off-site, online, and off-line. Regulations imposed as a result of the Health Insurance Portability and Accountability Act (HIPAA) will prompt the need for more secure storage modes. Satisfying that need will cost money, time, and staff. Storage giant EMC might ease the pain, at least a bit. The company has released a new platform called Centera that company execs believe will attract customers from the medical imaging community, among others.

What sets Centera apart from other storage technologies, such as area networks (designed for a transactional environment) and network-addressed storage (designed for a collaborative environment) is Centera's dedication to fixed content--"unchanging data objects with long-term value," said Joe Tucci, CEO and president of EMC. Medical images are a prime example of fixed content, he said, because changes typically are not made to an image after it becomes part of the patient's medical record.

Centera takes advantage of this fixed status when archiving data. By running a 128-bit algorithm on the zeros and ones that create a digital file, Centera produces a unique "content-ticket" that acts as a document identifier and verification mechanism. Jim Rothnie, senior vice president and chief technology officer of EMC, cites major benefits of this new way of dealing with information as data immutability, archiving, and manageability.

Data immutability provides proof that stored data have not changed before their retrieval. This is achieved by rerunning the algorithm and comparing content addresses, Rothnie said. In regard to archiving, once a piece of information is stored, Centera recognizes it and refuses to allocate more space to the same data if another user attempts to store it again. That saves storage space. Manageability is enhanced because Centera stores data as though it were using containers, Rothnie said, rather than location.

"Data goes in, a claim check comes out," he said. "It's like valet parking."

According to Rothnie, the data inside Centera is not in the form of a "volume" (as in storage area networks) or files (as in network-addressed storage). Rather, it is in the form of a binary large object. As is the case with DICOM, Centera attaches descriptive metadata to the object, such as the patient name and data format. Centera does not yet support direct querying of metadata, Rothnie said, but the company plans to release an upgrade by the end of the year that allows this function.

EMC estimates that the fixed content market will be worth about $3 billion in 2003 and $10 billion in 2005. Centera is EMC's means for obtaining a piece of that pie. Since its formal introduction April 29, several companies in various industries have announced support for the new technology. Among them is PACS developer Amicas. The company has already incorporated Centera into its Web-based PACS, and will be installing the first system in May at Newton Wellesley Hospital, a 249-bed acute care facility in Boston. Existing Amicas users can also upgrade to Centera.

"As far as our customers are concerned, the Centera software is an automatic part of the next upgrade," said Hamid Tabatabaie, CEO of Amicas. "There is no charge, no hassle. Centera supports DICOM, and the software is shipping with the capability of storing DICOM or JPEG 2000. But customers will need to choose to buy the Centera storage hardware to use the Centera software."

Centera uses a new type of architecture: redundant array of independent nodes (RAIN). Each node is about the size of a pizza box and sports a Pentium processor with 640 GB of memory. Each Centera cabinet can support 32 nodes, which store 10 TB of data objects. The data in Centera are mirrored across nodes. Adding capacity is as simple as plugging in a new node or a new cabinet. Up to 16 cabinets can be networked to form a cluster. Up to seven clusters can be connected, providing 1 petabyte (the equivalent of 1024 TB) of mirrored storage. Centera also supports geographic separation of elements of data in a system, which allows for disaster recovery.

"What we get is reliability, self-manageability, and hyperscalability," Rothnie said. "Next year we will have a terabyte processor and the new technology will participate with the old. As far as cost effectiveness, for the systems shipping today, a 16-node system costs $100,000--about 2¢ per megabyte--including hardware and software."

Although it is currently the sole source for Centera hardware, EMC hopes other vendors will begin developing hardware to support the company's storage platform, acquired when EMC purchased Belgian firm FilePool in 2001. Part of EMC's market strategy is to become a middleware company, rather than a pure hardware provider.

"We are passing along savings to the customer," Rothnie said. "We show prices separately for hardware and software, and we will welcome other suppliers building the hardware. We built the hardware first because there was none readily available. It was a time-to-market question. But we will earn our money with revolutionary software."

The cost benefits that go with Centera could entice customers to sign up. A community hospital conducting about 100,000 studies per year would spend only about $40,000 using a Centera/Amicas solution, according to Tabatabaie. A traditional PACS approach would cost much more, he said.

"The cost of tape drives and staff alone would exceed $40,000," Tabatabaie said. "It's a price-point issue--our customers have been waiting for something less expensive than disk and better than tape."

Centera could provide a major boost to PACS, according to Tabatabaie, because it provides customers with data authenticity, high availability, and cost-effectiveness. Plus, it eliminates the need for customers to be storage administrators.

"A PACS administrator has to continually map new drives, add storage, and figure out how to move files around to accommodate new systems," Tabatabaie said. "That is a pain. You typically experience downtime when mounting new drives and when migrating data. The Amicas application and its handshaking with the Centera environment eliminate that."

Other PACS vendors have also stepped forward as EMC partners--Agfa, Kodak, and Merge Technologies. Aside from Amicas, however, no other vendor is currently shipping Centera-enabled PACS. And, of course, not all PACS vendors are leaping on the EMC bandwagon.

Stentor, which received $20 million in second-round funding from EMC, has not switched its iVault storage over to Centera--at least not yet.

"Basically, although we can't comment specifically on the blueprints for future products, we are encouraged that EMC is reducing its price point with Centera to be more consistent with the medical imaging industry," said Mark Reis, marketing communications manager for Stentor. "As our industry evolves, lower cost storage for large medical imaging data sets is going to be greatly needed."

InSite One, which uses an ASP model for storage, doesn't plan to switch to Centera anytime soon. According to CEO Dave Cook, the company can provide storage at less than half the cost of implementing Centera.

"We have a similar architecture, and we can provide it a lot less expensively, although we're not getting into the whole universe of fixed content like EMC," Cook said.

InSite One will continue to market its storage service as a standalone and through partners like Philips and Agfa. The mode of delivery will depend on what customers want, Cook said. For example, a customer can buy an Agfa system with InSite One storage or an Agfa system with a Centera backend. But Cook believes that the advantage for customers lies in purchasing a service rather than a product.

"You can't win in storage--you're going to run out of space, or buy too much, or the storage technology will become obsolete," he said. "If you go with Centera, you're putting all your eggs in that architecture."