The original Netgear Inc. IP SAN was constructed with components from Bell Microproducts Inc., Seagate Technology, Marvell Technology Group and Zetera Corp., and had space for two 3.5-inch Integrated Drive Electronics drives of any size that mirrored each other. Zetera's firmware also enabled partitioning of the drives and the ability to string multiple Storage Central arrays together. At the time, the company claimed it would be theoretically possible to scale the system up to 128 petabytes (PB), but the initial version scaled up to just 1 terabyte.
In order to support the number of video files needed for the project, as well as the compute power necessary for the heavy calculations and storage of their output, "The array … will be built out over time to a massive 1.4 PB -- enough storage capacity for a high-resolution digital photo of every human being in the world," according to a press release put out by MIT last week.
"There's actually a lot of potential applications for this," said Jon William Toigo, CEO of Toigo Productions, who has been the technology's biggest champion so far. "If MIT's use goes well, it could really establish Zetera as an up and comer in the industry."
The lab affiliated with Toigo's company, TPI Technologies Inc., has been testing the Netgear equipment against other small SANs. Toigo notes the company is not paying for the research.
Rather, Toigo says he is enthused about a product that allows users to build their own SAN infrastructures from cheap commodity hardware. If it's done right, Toigo said, the performance numbers have been shown in TPI's lab to be superior to proprietary arrays including Dell Inc.'s and EMC Corp.'s AX100 products.
The system does storage service functions like replication, mirroring and RAID striping via a process called IP multicasting in which a switch sends data to multiple IP-addressed disks simultaneously. This is a proposition much less expensive than purchasing software to perform the tasks on the major vendors' arrays.
It has been pointed out that Zetera uses a proprietary version of User Datagram Protocol (UDP), rather than TCP used by iSCSI, in order to boost performance without the need for TCP offload engine cards.
"It's true that how they use it is proprietary," Toigo said. "As proprietary as any other vendor's technology inside the array -- the scheme of getting data on to and off of the disk is because of their own internal development."
Toigo also pointed out that the Netgear SAN will now support an NT file system as an alternative to a SAN file system from Data Plow Inc., in order to offer more open source options.
Other analysts more skeptical
"They very well may be on to something, but the point is there is a heck of a lot more to becoming a commercial success than spouting a statistic or a protocol in an academic environment where it's OK to play with toys," said Steve Duplessie, founder and analyst with the Enterprise Strategy Group. "InfiniBand was going to revolutionize the world a few years ago -- and it was proven to be better technically, but the business side didn't work."
"There's no question that five years from now, significant chunks of the enterprise storage environment will have moved over to entirely new IP-based products," said Brad O'Neill, senior analyst with the Taneja Group. "But it's still in the very early stages, and we still don't know what the IP architecture is ultimately going to look like -- it's going to take at least five years before a lot of this radical stuff comes off the fringe."
If the system's scalability can successfully be demonstrated, bigger enterprise users processing large amounts of data or who want to get away from proprietary replication and mirroring technology could take an interest in it. However, since the system uses UDP rather than TCP/IP as a protocol, it does not support the security standard for IP, called IPsec. Thus, it may not find as much penetration into companies with sensitive data in distributed environments.
Meanwhile, though Zetera says its goal is "to make its network storage technology the de facto industry standard," but so far its only customer traction has been in the consumer and small to midsized business (SMB) markets, where its biggest attractions -- simplicity and cost-effectiveness -- are more of a priority. It remains to be seen whether the product can break into the mainstream.
"They need to answer a lot of security and reliability questions before being considered an enterprise product," O'Neill said. "There's also the matter of software integration."