New Supercomputer in Pittsburgh Can Discover Never-Seen Patterns in Large Data Sets
The Pittsburgh Supercomputing Center (PSC) is introducing its latest, one-of-a-kind supercomputer to the national community Friday.
This Sherlock isn’t like Mr. Holmes. It doesn’t involve a magnifying glass, long trench coat, or pipe, but it does use heavy-duty problem solving. PSC’s Sherlock is the latest combination of hardware and software that will allow researchers and industries to look at large-scale data sets and graphs and turn them into understandable information.
The official debut on February 1st will include speakers and a demonstration of what Sherlock can do. In the meantime, Strategic Application Director at Pittsburgh Supercomputing Center, Nick Nystrom, said Sherlock is very different from a typical computer.
“It has, first of all, very large shared memory and the architecture can support up to 512 terabytes,” Nystrom said. “It also has unique capabilities for very massive multi-threading, and what that means is you can run many threads of a program on each processor.”
Nystrom said only one or two threads can run at a time on a typical laptop or desktop, but Sherlock can handle 128 threads at a time. He said that makes the computer spend less time drawing information from memory and more time doing calculations quickly. Sherlock can also find hidden relationships in large datasets.
Nystrom said Sherlock allows industries to look at their data on a large-scale for the first time.
“Understanding how social networks work, to cyber security, there are researchers looking at understanding protein pathways in various types of tumors, there are people looking at telephone networks and anomalies and understanding how burst patterns occur and what constitutes normal and irregular usage,” Nystrom said.