Search engine consolidates with NAS gateways
31 Jan 2006 | SearchVirtualStorage.com
Over the past year, online shopping search engine Shopzilla has moved to consolidate its backup, SAN and data warehousing systems -- and they've done so in each case by fronting the systems with Onstor Inc.'s NAS gateways.
The moves began last June, when the company tired of trying to use storage spread over 150 servers to run its proprietary search engine databases, and implemented a solid state disk SAN from Texas Memory Systems Inc. instead. To further consolidate the system, each of the Web site's total of 1,800-plus Linux production servers began to access the SAN through an OnStor Bobcat gateway.
"Suddenly, not only did we increase the speed of the disk on the back end, but now 300 clients have easy access to that data at a time [through the OnStor gateway], which they need," said Shopzilla's director if IT, Burzin Engineer. "The way we make money online is by applying user data in real-time to change our Web site."
In a typical infrastructure, application servers use a file system to organize data, and then SAN storage converts that data to a block-based system on the back end. The application server also becomes the "librarian" for access to the data -- all other applications looking to access the data have to go through it, which slows down the system.
What the OnStor gateway does is move the conversion of data and access point out of individual servers, allowing many applications to access data at the same time, and access it faster.
Engineer said Shopzilla was so pleased with the performance they saw from the OnStor gateway with their SAN that they chose a NAS/SAN combination from OnStor and 3Par Data, Inc. over expanding the EMC Corp. SAN and BlueArc Corp. NAS system they already had in place. The EMC and BlueArc equipment will continue to be used in his environment, Engineer said, but "in the future, [the OnStor system] is the direction we would like to go in.
"We were really unhappy with the costs associated with both EMC and BlueArc," Engineer said. "With the EMC sometimes you had to just pay to find out what you had, and the BlueArc Titans we were using came with their own disk."
Engineer said the OnStor is a gateway only -- meaning it can front any kind of disk. He went with SAN storage from 3Par because it was cheapest, but said he liked the idea of being able to change the disk behind the OnStor gateway on a whim.
"With the BlueArc, if you do something like that, they take away your warranty," he said. "You have to use the disk that comes with the Titan."
In the end, Engineer said, his new NAS/SAN system, including two OnStor gateway boxes for failover purposes, cost him less than two BlueArc boxes alone -- while the OnStor gateway gave his server clients even faster access to his SAN data, and consolidated the architecture into a unified system.
Last month, Engineer said, the company encountered yet another storage bottleneck.
"Our backups were becoming nearly impossible with our Veritas NetBackup software and ADIC tape robot," Engineer said. "We decided to put the SATABeast array in front of the tape backup to remedy that problem."
Engineer said it was a no-brainer to front the new disk backup with an OnStor gateway.
"We didn't need to go buy a high-end device, either," he said. "We got a Bobcat 2220 system on the cheap, and it works beautifully."
The new disk backup system came just in the nick of time, Engineer said, as the holiday season swung into gear. "December is our month to really shine," he said. "We were typically backing up between 300 and 700 gigabytes per day -- backing that up to tape would have been an impossibility."
While Engineer says he's pleased as punch with the OnStor system, there is one feature on his wish list -- right now, the Bobcat provides his servers read-only access to data, and he'd like for it to eventually allow the same multipath access for writing to his disk as well.
"That would allow us to really fine-tune our website data in response to searches," he said.