Skip directly to content

Sherlock unravels complex networks of data

NSF Award:

Enabling Productive, High-Performance Data Analytics  (Carnegie-Mellon University)

State:
Congressional Districts:
Research Areas:

Increasingly, research problems in security, medicine, public health and social dynamics require the ability to understand how large networks operate and change with time. Sherlock, a computing resource housed at the Pittsburgh Supercomputing Center (PSC), analyzes such complex networks of data without the memory lag inherent in other computers.

The computing system can pick apart and understand large chunks of a network at once, without having to split the problem into many pieces and work on them in isolation. For these types of networks, that means far fewer trips back to the stored data to assemble its calculations, surmounting the "memory wall" that previously slowed such computations. In addition, PSC staff customized Sherlock, a universal RDF integration knowledge graph analytics appliance (uRiKA), to give it a broader application for academic problems requiring massive memory to solve.

With 128 hardware threads per processor (32 times that of a typical supercomputer) Sherlock can "follow" multiple leads at once during a network investigation. Combined with the system's terabyte of shared memory--equal to the RAM of roughly a thousand new iPads--Sherlock offers the ability to find, for example, hidden sources of computer malware on the Web, new targets for cancer therapy and unexpected points at which epidemics may be halted.

Image

  • the sherlock supercomputer
Sherlock
PSC

Recent Award Highlights

a patient-specific geometric model and mesh of human veins with implanted medical device

Virtual human bodies guide vascular therapy

Patient-specific models test medical device effectiveness

Research Areas: Computing, Engineering Locations: Pennsylvania
a sample screen from quantum's braille tutor

Braille Learning on Demand

Artificial intelligence software helps the blind and visually impaired learn Braille

Research Areas: Engineering, Computing Locations: Pennsylvania