Premium
Introduction to special issue on selected papers from Energy Efficiency in Large‐Scale Distributed Systems 2013 conference
Author(s) -
Pierson JeanMarc,
Dittmann Lars,
Da Costa Georges
Publication year - 2014
Publication title -
concurrency and computation: practice and experience
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.309
H-Index - 67
eISSN - 1532-0634
pISSN - 1532-0626
DOI - 10.1002/cpe.3339
Subject(s) - computer science , scope (computer science) , efficient energy use , energy consumption , concurrency , scale (ratio) , distributed computing , operations research , industrial engineering , engineering , electrical engineering , physics , quantum mechanics , programming language
We are happy to present you this special issue dedicated to Energy Efficiency in Large-Scale Distributed Systems conference (EE-LSDS2013) of Concurrency and Communications: Practice and Experience. This special issue contains three papers selected from the EE-LSDS 2013 conference. The conference constituted the main outcome of the COST Action IC0804. The European Cooperation in Science and Technology (COST) Action is a 4-year funding scheme in European research framework aimed at developing researchers’ networks (www.cost.eu). The COST Action IC0804 (www.cost804.org) lasted from May 2009 and finished in May 2013. It investigated energy efficiency in large-scale distributed systems. For this special issue, we accepted three top contributions after a thorough review procedure. All the contributions relate to the scope of the journal and cover the High Performance Computing part of the activities presented at the conference. HPC system administrators are usually trying to tune their system to extract its peak or maximum performance. Energy consumed is then a second zone citizen and only taken into account when electricity bills arrive. Yet, energy and performance are not always opposite in HPC systems and can be studied and optimized together. More and more research works are investigating this link. The first step is to model the impact of green leverages such as virtual machines and migration, Dynamic Voltage Frequency Scaling (DVFS) or more generally hardware resource tuning on performance and energy. In [1], authors model virtualized computing nodes. They show that using such a model, energy consumption can be estimated allowing to make further optimization in the tuning, the scheduling or for accounting purpose. Then using such model, a runtime is needed to implement the decisions. In [2], authors use the fact that a large part of HPC applications runs tasks with precedence relations between them. They demonstrate on a matrix runtime that using the tasks configurations can save energy without impacting performance by smartly managing idle times in hybrid environments CPU/GPU. In a more generic way, authors in [3] show how they can apply green leverages at runtime with a small impact on performance by detecting with heuristics the type of workload on resources, and adjusting the resources accordingly (using DVFS, turning off components, etc.). We believe the readers will enjoy the papers in this special issue and will find them useful. The guest editors wish to thank the authors of all submitted manuscripts, without whom this special issue would not have been possible. They want also to outline the hard work of the reviewers who provided a thorough evaluation of the submitted manuscripts through several iterations. We greatly appreciated the support and assistance of the Editor-in-Chief, Geoffrey Fox.