Skip to Content
Find More Like This
Return to Search

Identifying Computational Operations Based on Power Measurements

Lawrence Berkeley National Laboratory

Contact LBL About This Technology

Technology Marketing Summary

Berkeley Lab researchers Sean Peisert and Charles McParland have developed a technology for power efficiency adjustments and for monitoring use or unauthorized use of computing systems, including supercomputers and large computing centers. By using power data, as opposed to data provided by the computing environment itself, the technology collects the data non-invasively.

Description

Berkeley Lab researchers Sean Peisert and Charles McParland have developed a technology for power efficiency adjustments and for monitoring use or unauthorized use of computing systems, including supercomputers and large computing centers. By using power data, as opposed to data provided by the computing environment itself, the technology collects the data non-invasively.

 

Highly sensitive power measurement devices analyze, at a very high level of granularity and resolution, the power loads drawn by HPC systems, including, but not limited to CPUs, GPUs, and I/0 (e.g., hard disk, SSD). The technology enables “fingerprinting” or inferring of computation by examining patterns in power use of processors and I/0 systems through micro phasor measurement units and power quality meters.

 

Early approaches captured MPI function calls via the IPM performance monitoring library to describe patterns of communication between an HPC system and nodes. However, they imposed high enough computational overhead that they were not used in practice. The new Berkeley Lab technology based on power analysis is much less invasive and, therefore, much more likely to be put into regular use.

Benefits
  • Non-invasive monitoring with low risk of detection
 
Applications and Industries
  • Signals intelligence for computers, including supercomputers
  • Misuse identification of computing systems
  • Relevant to large, power-consuming data centers, i.e., Amazon, Google, IBM, etc. 
 
Technology Status
Technology IDDevelopment StageAvailabilityPublishedLast Updated
2016-053ProposedAvailable02/04/201702/04/2017

Contact LBL About This Technology

To: Suzanne Storar<ipo@lbl.gov>