Carroll and his team needed the array because the Institute is creating data at the rate of up to five terabytes each day, well in advance of planned storage growth.
He already has 15 terabytes of live data plus another 150 terabytes tucked away in archives, but the Sun/Oracle array he was using as storage for researchers’ data and administrative material for the Institute was struggling to cope.
“We outgrew its capacity and exceeded its controller’s IOPs,” explains Toby O’Brien, a Senior Systems Engineer who works with Carroll.
O’Brien was given the job of finding a replacement and settled on the HDS device, citing the new machine’s possession of “more spindles and ports” than its predecessor.
Those technical features made the purchase worthwhile, but also raised another issue: the Institute was a Sun shop, uses Sun/Oracle hardware and software extensively and wondered what it would mean to introduce a single new device from a new vendor.
“It always creates a little bit of friction between vendors,” Carroll told SearchCIO ANZ. “They say: ‘You are ruining the pure stack, man! What are you doing?’”
Carroll resolved that even though the new array was from a new vendor, he could manage both the device and the additional relationship.
“It’s just disk,” he says. “And managing a contract is managing a contract. If you have one, two or ten ... it’s all the same.”
Carroll also has users to satisfy, and is not convinced that any storage vendor understands the things his will do to an array.
“We have to be agile. Researchers show up with new instruments,” and these devices create lot of data. “We have to be agile enough to suck it up. All our experience suggests that people who forecast storage growth are in the enterprise, not research institutions. We have extremely different needs to the enterprise, although we do buy these arrays because they have the features a generic enterprise looks for.”