RapidScale consists of a global namespace and NAS hardware nodes designed for scale-out and high-performance computing (HPC) applications. Rackable also acquired the rights to Terrascale's Distributed Parity Engine in April 2007 from the former shareholders of the company. Distributed Parity Engine ensures that the global file system will stay available during the failure or outage of a storage target.
The RapidScale patents cover management of local client cache buffers in a clustered computer environment, methods for retrieving and modifying data elements on a shared medium, and methods and systems for network storage device failure protection and recovery. Rackable paid $38 million to acquire Terrascale.
The to-be-announced storage partnership could come by the end of the third quarter, according to Barrenechea, adding, "But at this juncture, we believe RapidScale is no longer core to our strategy." During the call, Barrenechea said the "most attractive" alternative would be to sell the RapidScale business to a partner that Rackable can maintain a relationship with.
Rackable has retained the bank Strategic Advisory Services International to seek strategic alternatives for the RapidScale product. Rackable general counsel Maurice Leibenstern said options include a deal with another storage company or a private equity firm. "The bottom line is that it would take full-time effort and investment dollars to bring [RapidScale] to fruition the way it should be," he said.
Some analysts said they werent' surprised to see Rackable dump RapidScale. "Rackable is fundamentally a hardware company, and clustered file systems, while they have a hardware component in scale-out systems, are essentially software," said Arun Taneja of the Taneja Group, who advised Terrascale prior to its acquisition by Rackable. "Many hardware companies have gone into the software business looking to increase gross margins and find it's a very different business."
Other analysts speculate that Terrascale's architecture is better suited to large files in HPC environments, rather than serving the small files that tend to pop up in Web 2.0 data centers with large amounts of online transaction processing. "Everything I've heard about the product is that it really does work," said Data Mobility Group analyst Robin Harris. "It just may not be a particularly good fit for Rackable's customer base.
Harris said that Barranechea's reference to "fault-tolerant enterprise workload" sounds like IBM's soon-to-be relaunched XIV [Ltd.] block-level clustering system. "If this is a deal with IBM," he said, "it would be a great way for IBM to accelerate the market penetration of XIV through a partner deploying it in a controlled environment."
Leibenstern would neither confirm nor deny a storage partnership with IBM, offer any further clues as to who the partner would be, or say exactly when details of the deal will be announced.
Storage has accounted for 13% of Rackable's revenues so far this year. However, on the August 4 call, chief financial officer James Wheat said, "We concluded that this business was not likely to meet our future expectations. Accordingly, we are considering strategic alternatives to enable us to exit the business."