Component Object Based Single System Image Middleware for Metacomputer Implementation of Genetic Programming on Clusters

Authors:
Ivan Tanev;Takashi Uozumi;Dauren Akhmetov
Affiliations:
-;-;-
Venue:
ICCS '01 Proceedings of the International Conference on Computational Sciences-Part I
Year:
2001

Citing 5
Cited 1

System identification: theory for the user

System identification: theory for the user
Parallel genetic programming and its application to trading model induction

Parallel Computing
High Performance Cluster Computing: Architectures and Systems

High Performance Cluster Computing: Architectures and Systems
Genetic Programming III: Darwinian Invention & Problem Solving

Genetic Programming III: Darwinian Invention & Problem Solving
Speeding up genetic programming: a parallel BSP implementation

GECCO '96 Proceedings of the 1st annual conference on Genetic and evolutionary computation

Component Object Based Single System Image for Dependable Implementation of Genetic Programming on Clusters

Cluster Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

We present a distributed component-object model (DCOM) based single system image middleware (SSIM) for metacomputer implementation of genetic programming (MIGP). MIGP is aimed to significantly improve the computational performance of genetic programming (GP) exploiting the inherent parallelism in GP among the evaluation of individuals. It runs on cost-effective clusters of commodity, non-dedicated, heterogeneous workstations. Developed SSIM represents these workstations as a unified virtual resource and addresses the issues of locating and allocating the physical resources, communicating between the entities of MIGP, scheduling and load balance. Adopting DCOM as a communicating paradigm offers the benefits of software platformand network protocol neutrality of proposed implementation; and the generic support for the issues of locating, allocating and security of the distributed entities of MIGP. Presented results of experimentally obtained speedup characteristics show close to linear speedup of MIGP for solving the time series identification problem on cluster of 10 W2K workstations.