"Part of what we do is clinical proteomics," Mallick explains. "The goal is to be able to take a drop of blood from a patient, measure it in extreme detail, and based on what we find, diagnose if someone has cancer or not."

It takes one hour to examine a single drop of a patient's blood in the mass spectrometer at Cedars-Sinai, and the result creates about 50GB of time-series data. All told, this process generates about 1TB of data per day.

"We have a high-performance computing cluster -- I think it's rated 367th among the top 500 supercomputers -- which we use to do sophisticated computational analysis on all the data we collect," Mallick says. "We want to discover patterns that differentiate a patient group from another. If you can identify who is likely to respond [to a specific therapy], you've saved a lot of lives."

Managing that data had been a problem, however. Many of the solutions Cedars-Sinai tried used tapes as their last-tier medium, which made computational analysis and pattern searches unacceptably slow. Other solutions were cumbersome and made moving data across tiers exceedingly complicated. Adding capacity required significant technical prowess and occasionally forced Cedars-Sinai to break its blood archive in smaller sections. It was time for a change.

"We wanted something that was easy to use, highly available, could scale very easily, and could also handle the performance requirements," Mallick explains.