Futures market bets big on start-up

Chicago Mercantile Exchange IT leader finds futures market banks on storage start-up Copan.

At the Chicago Mercantile Exchange, Craig Taylor’s storage needs had grown exponentially, from 4T bytes two and a half years ago to more than 180T bytes today. Backing up all that data had overloaded his StorageTek tape libraries, caused failures with the robotic arm and resulted in one in four backups failing.

“We had performance issues, and the backups became fragmented over time,” says Taylor, associate director of open systems for the exchange.

Taylor set out to find a product that would solve this problem and let him back up the 4.2 million futures and options contracts the exchange processes each day.

Starting in 2003, Taylor began evaluating traditional storage gear. In 2004, he chose a new technology from Copan Systems, a storage start-up in Longmont, Colo., that was founded in 2002 and shipped its first product in 2003.

CME made an initial purchase in late 2004 and additional purchases as it rolled out the gear in 2005.

“We looked at IBM, EMC, StorageTek, Sun Serial ATA technologies,” Taylor says. “When it came to price and functionality, Copan’s Revolution 200T rose to the top of all of them.” Copan’s Revolution 200T is a virtual tape library that uses  disk drives instead of more-expensive drives. A virtual tape library emulates the actions and processes of traditional tape libraries.

MAID to order

Because his organization buys conservatively, purchasing equipment from a start-up was a concern to Taylor and his bosses.

“We looked at the risk of using a new technology like Copan’s and did a risk assessment,” Taylor says. “We decided that there was a bigger risk if we didn’t make changes than if we did.”

Taylor was looking for gear that used Serial ATA drives. The duty cycle – when the disks are spinning – of Serial ATA drives is between 25% and 50%. The products from the other companies Taylor was looking at spun their Serial ATA disks 100% of the time, meaning that there would be more frequent drive failures.

In contrast, the Revolution 200T uses a technology called Massive Array of Idle Disks (MAID): When data is needed, the disks are spun and data is retrieved. When there is no activity required, the disks sit idle, waiting for the next disk access. Using this technology saves companies power and heat consumption, and saves customers money.

In addition, because the disks are idle much of the time, the chances that they will fail are less than if they spin continuously. Copan regularly checks the operation of the disks using what it calls Disk Aerobics, which exercises each of the disks and monitors them to make sure they are healthy.

Only a handful of vendors produce disks using MAID technology, making Taylor’s adoption of it more perilous. Among them are Asaca and Exavio, two companies that specialize in the video broadcast market, and Nexsan, which recently introduced its SATAbeast product.

While Taylor won’t discuss what he paid for the Copan system, he estimates he will save $30,000 a year over three years in power costs alone. According to Copan, its Revolution 200T costs about $3.50 a gigabyte, whereas the cost of other virtual tape libraries ranges from $8 to $20 a gigabyte.

Tiered terabytes

Taylor added the 320T-byte Revolution 200T to his storage-area network () as a fourth tier of storage. His Tier 1 data – the applications and databases that run the exchange’s everyday business – resides on EMC Symmetrix DMX storage. Tier 2 is for older Hitachi Data Systems 9960 arrays, where less business-critical production data is stored. Data that needs to be kept a long time is moved to Tier 3, which itself is undergoing migration from HP 2200MX magneto-optical storage to EMC’s Centera content-addressable storage.

Finally, data stored in Tiers 1 and 2 is backed up to the Tier 4 Copan system. Adding Copan to his SAN freed up the old StorageTek PowerHorn tape libraries for use in offsite disaster recovery and storage. However, those libraries are now being switched in favor of IBM 3584 tape libraries.

“The Copan system is for write-once, read-almost-never data,” Taylor says. That’s why the spin-up-when-needed technology doesn’t bother him. He doesn’t need to get to it that often and when he does, the slower access to the data doesn’t affect his operations. “Waiting 15 seconds for data you almost never need is not a problem,” he says.

Performance gains

Even with the slower access time of MAID technology, Taylor is able to complete his backups faster than with the previous StorageTek-based backup. “Copan testing showed up to a 228% gain in back-up and restore times,” Taylor says. “We never run past our back-up window any more.”

In addition, he hasn’t suffered any more of the failures associated with tape-based media and, in using the Revolution 200T for the past year, the operation has not experienced any drive failures.

Taylor uses NetBackup software from Symantec to protect his data. No changes other than swapping out the StorageTek libraries for the Copan array were needed to establish a new back-up operation and get Taylor’s backups moving again.

As for going with a start-up vendor and new technology, Taylor doesn’t hesitate. “We still have the tape in place, and if Copan were to go away, we could resume backing up to tape,” he says.

This year, Taylor plans to evaluate the use of another Copan Revolution 200T to replicate data for disaster recovery and for archiving his compliance-based data.