Bringing CAS to the masses

Back in 2001 when EMC bought Belgian startup FilePool, only a few people knew what CAS (content addressable storage) was. Even fewer people knew that FilePool's Paul Carpentier was the mind behind CAS.

Nevertheless, CAS was (and still is) a major revolution in storage because it identifies, I believe for the first time ever, a separated data custodian role. In essence, it's a system that provides reliable, centralized archival services to a variety of clients.

Here's a quick recap of how CAS works: A client application delivers a file to a CAS system and receives in return a token that uniquely identifies the file but is also a verification key that certifies the integrity of the data content.

Because of its characteristics, a CAS system is a simple and effective repository for long-term archival. It's especially good for fixed-content files that either don't change at all or seldom change over their lifetime, such as the digital image of a contract.

Back to the news at hand: About a year after acquiring FilePool, EMC announced Centera, a new product line based on CAS. Since then, EMC has made that technology one of the cornerstones of its ILM (information lifecycle management) and compliance strategies, including eDiscovery.

If you've ever thought that -- however interesting it may be -- Centera is a bit too rich for your blood and your budget, you're not alone. In fact, CAS inventor Carpentier has founded a new company, Caringo, to prove that point.

Paul Carpentier is CTO of the startup Caringo, and President Jonathan Ring and CEO Mark Goros round out the management team. In case you missed it, the name of the company comes from the initial letters of the three managers' last names.

At the end of May, Caringo announced its first product, CAStor, a software implementation of CAS. "We deliver on a USB key a software product that runs on x86 commodity hardware," Goros says. "Sixty seconds after you boot from that USB key, you have a running CAStor node."

Adding a second node is as easy as booting a second machine: CAStor expands automatically as you add more servers and will redistribute its data content to take advantage of the new nodes.

CAStor can scale to hundreds or thousands of nodes linked by Gigabit Ethernet with little or no human manipulation. "It manages itself, it configures itself, and you can add nodes while it's running," Goros remarks.

In addition to the CAS management code, CAStor includes a customized version of Linux. "We have removed anything that was not needed, including the possibility of remote access," Carpentier explains. "Our application turns a general-purpose server into a single-minded appliance that works together with others to preserve the integrity of fixed content files."

When I ask to see a screenshot of its management GUI, Carpentier has a simple answer. "That would be a problem," he says. "We don't have a management GUI."

"There is no centralized access, and no gateway. We use a subset of HTTP 1.1 [to communicate with clients]," Carpentier continues, explaining that a client can contact any node and that the request to add a new file or to retrieve existing files will be broadcast to every node in the system. The node that wins the bid gets the job, Carpentier explains.