Load Balance Protocol of Cluster on Grid: Pervasive Maximum Algorithmic Parallelism

Description

Pervasive computing software helps to manage information and reduce the complexity of available computing resources in a timely manner. On one hand, a grid is a resource whose information is changing anytime and anywhere, where the availability of CPUs is only informed in run time when a cluster asks to the grid. On the other hand, a cluster is often implemented in a static way, which assumes some particular parallel architecture and the number of CPUs. Even when a cluster can consume maximum available CPU resources, if a cluster is implemented assuming the number of CPUs, a cluster could not run using more than this number of CPUs. Our mechanism of load balance protocol of cluster provides a dynamic way of implementing a cluster that can consume maximum available CPU resources on run time. In order to do so, we propose two mechanisms: to provide yet another (graphical) parallel language to describe clusters and to provide the protocols between a cluster on a grid. While many parallel languages resolve parallel architecture dependencies in compile time, our parallel language resolve parallel architecture dependencies in run time. Our load balance protocol bases on this (graphical) parallel language and it provides implementation of them.

Details 詳細情報について

Report a problem

Back to top