TR-2015-04
MEANTIME: Achieving Both Minimal Energy and Timeliness with Approximate Computing
Anne Farrell; Henry Hoffmann. 15 May, 2015.
Communicated by Henry Hoffmann.
Abstract
Energy efficiency and timeliness (i.e., predictable
job latency) are two central concerns for real-time systems. While
both are essential, they are opposing: hard timing guarantees require
conservative resource allocation while energy minimization
requires aggressively releasing resources and occasionally violating
timing constraints. Recent work on approximate computing,
however, opens up a new dimension of optimization: application
accuracy. In this paper, we use approximate computing to achieve
both hard timing guarantees and energy efficiency. Specifically,
we propose MEANTIME: a runtime control system that delivers
hard real-time latency guarantees and energy-minimal resource
usage by sacrificing a small amount of application accuracy.
We test MEANTIME on a real Linux/x86 system with seven
applications. Overall, we find that MEANTIME never violates
real-time deadlines and sacrifices a small amount (typically less
than 2%) of accuracy yet consumes only 68% of the energy of a
conservative, full accuracy approach.
Original Document
The original document is available in PDF (uploaded 15 May, 2015 by
Henry Hoffmann).
