Where academic tradition
meets the exciting future

You are here: TUCS > PUBLICATIONS > Publication Search > Analysis of Power Management S...

Analysis of Power Management Strategies for a Single-Chip Cloud Computer

Robert Eksten, Kameswar Rao Vaddina, Pasi Liljeberg, Juha Plosila, Analysis of Power Management Strategies for a Single-Chip Cloud Computer. In: Stefan Lankes, Carsten Clauss (Eds.), Proceedings of the Many-core Applications Research Community (MARC) Symposium at RWTH Aachen University, 32–37, Chair for Operating Systems, RWTH Aachen University , 2012.

Abstract:

The 48-core SCC allows us to test and analyze the efficiency of power management algorithms in a multicore environment. In this paper we explain how we implemented these algorithms on the Intel SCC and how they performed in comparison to each other. Two of the algorithms, Core-Pool and PAST, were introduced in previous research articles and one algorithm, Derivate, was developed in this research project. The benchmark chosen to test these algorithms was a NASA Advanced Supercomputing benchmark that was imported to the platform. We analyzed the algorithms according to the execution time of the benchmark and the average power consumption while the test was running. The algorithms were compared against each other and reference values while having the benchmark running at constant power level. Through this analysis, we were able to determine that Core-Pool gives the best results in power efficiency with PAST close behind with almost as promising results. Derivate on the other hand proved to be still lacking, and in need of redesign in order to provide results that are comparable to the other power management algorithms. Index Terms—power management algorithms, the single-chip cloud computer, multicore power management, dynamic voltage and frequency scaling.

Files:

Full publication in PDF-format

BibTeX entry:

@INPROCEEDINGS{inpEkVaLiPl12a,
  title = {Analysis of Power Management Strategies for a Single-Chip Cloud Computer},
  booktitle = {Proceedings of the Many-core Applications Research Community (MARC) Symposium at RWTH Aachen University},
  author = {Eksten, Robert and Vaddina, Kameswar Rao and Liljeberg, Pasi and Plosila, Juha},
  editor = {Lankes, Stefan and Clauss, Carsten},
  publisher = {Chair for Operating Systems, RWTH Aachen University },
  pages = {32–37},
  year = {2012},
  keywords = {power management algorithms, the single-chip cloud computer, multicore power management, dynamic voltage and frequency scaling},
}

Belongs to TUCS Research Unit(s): Embedded Computer and Electronic Systems (ECES)

Edit publication