IBM developing largest ever storage array

IBM is building a hard drive of sorts that's 10 times bigger than any other built before it. It has a capacity of 120 petabytes, or 120 million gigabytes, and is made up of 200,000 hard drives hooked up together. IBM's data storge group is working on the project at the hardware giant's Almaden, CA research lab.

It can store about one trillion files and would allow for more powerful simulations of complex systems, such as those used to model the weather and climate. The megaserver is being built for an unnamed client who is building a supercomputer for detailed simulations of real-world phenomena.

Lessons learned during its creation could lead to similar systems for commercial users, said director of storage research at IBM, Bruce Hillsberg, who is also heading up the project. He added that in a few years cloud storage systems may have many similarities to this system.

IBM had to develop a number of new hardware and software techniques in order to build this system. The stacked disks in the racks are cooled by circulating water rather than fans, for example. When a disk fails, the system pulls redundant data from other drives and slowly writes it back to a replacement. If more disks fail, the rewrite process speeds up to reduce the possibility of permanent data loss. According to Hillsberg, this results in a system that should not lose any data for a million years without a compromise in performance.

The largest current storage arrays are about 15 petabytes in size. [via MIT Technology Review]

By Electronista Staff