IBM’s data storage team is building a 120 petabyte drive array which consists of 200,000 normal hard drives working together. The client has not been named but what is known is that the client needs the storage for simulations of real-world phenomena like weather.
The 120 petabyte array uses specially modified drawers to hold more hard drives inside a standard data center cabinet. To keep the drives cool IBM opted for a liquid cooling system that is more reliable than standard fans as well as more energy efficient. To speed up data access times IBM will be using its General Parallel File System (GPFS). GPFS spreads data across multiple drives allowing many files to be read or written at the same time. The GPFS file system allows supercomputers and data clusters to keep track of where files are without having to scan the entire system looking for individual files. To keep track of all the files as well as their attributes on the 120 petabyte array the GPFS dedicates two petabytes to this job; which starts to make a 2GB page file on Windows look downright reasonable.
Source : Tom’s Hardware