In-Depth

Avamar’s Value Now Patently Obvious

Finally, after nearly six years, Avamar Technologies has received the third of the 22 patents it has applied for.

• By Jon William Toigo
• 01/27/2005

Technology patents constitute, more than anything else, bragging rights for the owner. They allow the recipient vendor’s marketing guys to argue that their company’s scientists are better than their competitors’ scientists. Too often, however, a patent award simply means that one application got the green light before a nearly identical one filed by another party. The first gets to brag, while the other eats crow. It’s been that way since Elisha Grey took Alexander Graham Bell to court over who had legitimate claim to the title of “father of the telephone,” and it hasn’t changed much since.

As a practical matter, patents have little to do with product value. Having a patented technology says absolutely nothing about the implementation of that technology in a product nor anything about the product’s actual business value. That’s why, when I get a press release across the transom about a patent just awarded to storage vendor XYZ, I tend to direct it to the round file.

However, when I received the news about Avamar Technologies’ recent patent award, I decided to call the Irvine, CA-based company. Why the different tact with Avamar?

Partly because I knew this company when they were just a smile on mommy’s face that daddy didn’t understand. As a start-up, they were full of energy and dead set on creating a technology for data archiving that would radically improve upon traditional modalities (such as tape) with an emphasis on squeezing out cost and improving efficiency.

This was back in the late 1990s, when no one cared a lick about costs or efficiency. Everyone knew about the problems with tape, but they responded usually by throwing more tape at the problem, or by purchasing expensive mirroring technologies that made them the towel boys to EMC, IBM, et al. When I traveled to Avamar to hear their pitch for a real, business savvy alternative, I liked their attitude.

My impression climbed even higher when they put Kevin Daly to work as CEO. Kevin is perhaps one of the most beloved and respected guys in the business. He is the father, mentor or grandpappy that everyone in my generation wished we had had. Having built up tape-library vendor ATL from scratch, he subsequently traveled with ATL into the arms of then-tape-powerhouse Quantum Corporation. He labored at Quantum to make their ATL division successful as CTO, then, in the early 2000’s left for the opportunity represented by Avamar.

And why not? Avamar Technologies was planning to revolutionize the world of archiving. They sought to create a hardware-independent technology that would enable you to shrink archival data to a fraction of its size and to posit it on an infinitely scalable set of independent nodes. Not a cluster, mind you (those are painfully difficult for most folks to deploy and maintain), but a new/old technology they called redundant array of independent nodes (RAIN). Just as RAID had revolutionized the realm of disk in the late 1970s, RAIN extended the model across multiple arrays. Such a platform would enable you to lose an entire array without losing data, a very important attribute of an archiving system that would need to handle data for a very long time.

There is no doubt that similar ideas were being pursued elsewhere, of course. At the last Storage Networking World, I had a chat with folks from Rocksoft, founded in 2000 in Adelaide, Australia, with offices in San Jose, CA and Dallas, TX, and with Data Domain, founded in Palo Alto, CA in 2001. Booth reps for both companies seemed too preoccupied with Avamar to tell me about their own wares. Each claimed that some part of Avamar’s technology had trodden on their intellectual property, and that they, not Avamar, should be regarded as the top dog of tape-archive replacement.

Rory Bolt, Avamar’s CTO, says that he knows the rant. He offers that independent like-minded individuals can reach similar technological conclusions. Avamar’s patent applications, he notes, disclosed prior, similar technology wherever it existed. In the end, however, Avamar’s patent applications got the nod and the company got the bragging rights.

To date, the company has a patent on its own “hash file system,” which, according to Bolt, is a key enabler of a form of content addressable storage (CAS) in which object properties determine the node address where contents will be stored. Unlike EMC Centera, there is no proprietary hardware involved. The solution is purely software-based (though the company offers its own hardware platforms to go with the software, if one so desires).

Another patent covers “sticky byte factoring”—a method for breaking up data into 10 to 20K chunks that comprise variable size objects, each with its own unique address. Bolt says that the company’s method for finding natural breakpoints in a file, and for shifting data in a file when changes are made, are part of this important patent. The methodology eliminates the need for a centralized database of file metadata, enabling instead the data to carry around all of its descriptive information.

The third patent, just awarded, goes by the Star Trek-sounding handle “multidimensional parity.” This is RAIN in a nutshell: a way to break up the universe of storage space into an array of hardware-agnostic nodes. “It very similar to grid computing architecture,” Bolt argues, “and enables parity data to be striped across all nodes in a system.” That way, the loss of a node, due to a catastrophic power supply failure or some other event, will not result in the loss of data.

Bolt agrees that patents do not tell the whole story of product value. He notes that Data Domain offers an improved data storage device, but not a better data-protection solution. “With their product, you need third-party software if you want to do what we do.”

Both Rocksoft and Data Domain, he notes, have cryptographic processes similar to his own company’s commonality factoring algorithm, which enables data compression. “But,” he adds, “they do not have our hash file system and do not scale the way that we do.”

Avamar’s solution, called Axion, is meeting with considerable success in the market, and was most recently adopted by Adaptec Corporation for handling the archival storage and restoration features of 12 TB of storage. It is increasingly the engine behind data protection services such as those offered by Arsenal Digital and MCI.

If you have experience with the Axion solution and want to share it with your peers, please contact me at jtoigo@intnet.net.