Researchers at IBM
’s Almaden, CA research facility are building a storage drive of monstrous proportions, a machination comprised of 200,000 hard disks that equals 120PB. So yes, it can house your entire digital music and HD video collection, as well as tens of thousands of copies of it.
The drive is designed to run the huge, complex simulations needed to research things like climate change and molecular study.
In order to make such a project work, the IBMers had to innovate. For example, the racks are densely populated with hard disks, creating a heat problem that the team solved by using water cooling instead of fans. Further, in order to prevent data loss, when a disk fails, the redundant data stored on other disks is slowly copied onto the failed disk’s replacement, which doesn’t really affect performance; if multiple nearby disks fail, that process is sped up to prevent any permanent data loss.
Purportedly, the 120PB drive will not lose any data or see an impact in performance for one million years.
Size, of course, is one thing, but speed is another. In order to boost performance, IBM is leveraging software, such as its own GPFS file system that it designed specifically for supercomputing applications.