Home         Authors   Papers   Year of conference   Themes   Organizations        To MES conference

Adoption of Genetic Algorithms for running in elastic compute environment concerning CAD applications
Authors
	Shlepnev A.A.
Date of publication
	2016
Abstract
	Genetic algorithms are widely used for optimization purposes in EDA CAD. This application usually requires a lot of computing power so resources of cloud providers could be engaged. GA are well known subject for parallel computing, but traditional approaches for parallelizing does not allow to take all advantages of Clouds, especially elasticity. In this article two ways of adoption of Genetic Algorithms for running in Cloud infrastructure are proposed. First way requires any kind of commercially available grid engine like LSF, OpenLAVA, SGE or Univa to run parallel parts of algorithm such as fitness function computing and uses file operations on common storage to provide synchronization among algorithm stages. Second way is bulding GA application using elastic architecture, similar to architecture of GRID engines for running parallel parts of algorithm, and using proposed «waiter» syncronization primitive, similar to tokens used in parallel dataflow computing systems for syncronization. This way allows to increase/reduce of computing workers on-the-fly depending on current workload to provide full elasticy of application and decrease latency of interprocess communications by hundreds times (comparing to the first way).
Keywords
	elastic computing, cloud computing, grid computing, genatic algorithm, optimization, parallel computing, infrastructure.
Library reference
	Shlepnev A.A. Adoption of Genetic Algorithms for running in elastic compute environment concerning CAD applications // Problems of Perspective Micro- and Nanoelectronic Systems Development - 2016. Proceedings / edited by A. Stempkovsky, Moscow, IPPM RAS, 2016. Part 2. P. 197-203.
URL of paper
	http://www.mes-conference.ru/data/year2016/pdf/D155.pdf