Tape storage isn’t what you’d call cutting-edge technology. Most of us make do with disk, and lust after the speeds of SSD. But tape is still useful when massive amounts of storage are needed, in part because of its low cost and power requirements. And it's being put to good use at an extreme scale in a new supercomputer.
According to Computerworld, the National Center for Supercomputing Applications (NCSA) is building a storage infrastructure consisting of 380 petabytes of magnetic tape capacity and 25 petabytes of disk storage. It's all to support the petaflop-scale Blue Waters supercomputer. The NCSA says it is building the system to "predict the behavior of complex biological systems, understand how the cosmos evolved after the Big Bang, design new materials at the atomic level, predict the behavior of hurricanes and tornadoes, and simulate complex engineered systems like the power distribution system and airplanes and automobiles."
The 25PB of disk will act as online storage for data that must be rapidly accessed, while the tape library is categorized as nearline, sort of a compromise between online storage and backup systems. With 380,000 AMD Opteron 6200 Series x86 processors (and 3,000 NVIDIA GPUs), the cluster will use 40Gbps Ethernet technology with aggregate throughput of up to a terabyte per second, Computerworld reported. The primary interconnect, however, is Cray's Gemini technology.
Spectra Logic, the company supplying the tape drives, said Blue Waters will take a couple of years to scale up to 380PB. Ultimately, it will become "one of the world’s largest active file repositories stored on tape media." The tape library itself will be capable of read/write speed of up to 2.2 petabytes per hour.
With some understatement, NCSA senior technical program manager Michelle Butler said the requirement to offload massive amounts of scientific data means "we needed a very large tape drive infrastructure." They’re also presumably going to need a heck of a lot of robots to manage it all.
