On the heels of China claiming the fastest supercomputer, Johns Hopkins is taking a bit of a different approach with its new machine. Rather than configure its new system for maximum FLOPS (floating operations per second), Johns Hopkins is configuring the machine to achieve the maximum amount of IOPS (I/O operations per second), says InfoWorld.
This approach will be well-suited to the kind of data mining-oriented scientific workloads processed by today’s supercomputers. Alexander Szalay, a computer scientist and astrophysicist at Johns Hopkins’ Institute for Data Intensive Engineering and Science, who is leading the project, dubbed Data-Scope, explained:
For the sciences, it is the I/O that is becoming the major bottleneck. People are running larger and larger simulations, and they take up so much memory, it is difficult to write the output to disk.
Data-Scope has received backing from the National Science Foundation in the form of a $2.1 million grant. When complete, Data-Scope will be able to handle five petabytes of data. Data-Scope will also allow Szalay, a host of researchers at Johns Hopkins and other institutions, including universities and national laboratories, to conduct research directly in the database, reports SpaceDaily.com.