Where academic tradition
meets the exciting future

You are here: TUCS > PUBLICATIONS > Publication Search > Energy and Power Management, M...

Energy and Power Management, Measurement and Analysis for Multi-Core Processors

Simon Holmbacka, Fredric Hällis, Wictor Lund, Sébastien Lafond, Johan Lilius, Energy and Power Management, Measurement and Analysis for Multi-Core Processors. TUCS Technical Reports 1117, TUCS, 2014.

Abstract:

This technical report introduces the current evolution of computing system and their power dissipation and energy consumption. It focuses on the power and energy consumption of microprocessor and presents in more details the two mechanisms, DVFS and DPM, available to increase the energy efficiency of computer systems. Different traditional approaches to evaluate the load on a processor are discussed and an extension of the notion of load taking into consideration Quality of Service of applications is proposed. Based on the extended load notion, the construction of power and performance models and a design of a dedicated benchmark, a run-time power manager is presented.

An evaluation of the presented benchmark and run-time power manager provides early results on the achievable power and energy savings and gives a comparison of the proposed approach with by default Completely Fair Linux scheduler and load balancer.

Finally we present two ways of conducting power measurements by using both internal and external power measurement devices. An open-hardware solution to measure power from any kind of chip (provided that the current feed pin are exposed) is presented together with an open-source logger software running on a low-cost Raspberry Pi platform. This provide one concrete example to create a cost efficient power tracing device without industrial scale manufacturing equipment.

Files:

Full publication in PDF-format

BibTeX entry:

@TECHREPORT{tHoHxLuLaLi14a,
  title = {Energy and Power Management, Measurement and Analysis for Multi-Core Processors},
  author = {Holmbacka, Simon and Hällis, Fredric and Lund, Wictor and Lafond, Sébastien and Lilius, Johan},
  number = {1117},
  series = {TUCS Technical Reports},
  publisher = {TUCS},
  year = {2014},
  keywords = { Power, Energy, Measurements, Multi-Core, Benchmark, Model-based run-time},
  ISBN = { 978-952-12-3087-5},
}

Belongs to TUCS Research Unit(s): Embedded Systems Laboratory (ESLAB)

Edit publication