|
Managing Parallel Computational Tasks on a Grid Environment A.I. Avetisyan, S.S. Gaissarian, D.A. Grushin, N.N. Kuzjurin, A.V. Shokurov Institute for System Programming of Russian Academy of Sciences Recent work in Grid environments enables applications to access computational resources in different and widely dispersed locations. Grid architecture includes homogeneous multiprocessor systems (e.g. clusters). Such systems may consist of hundreds or thousands of CPUs. We present an architecture and principles of hierarchical decentralized management of parallel computational tasks in such Grid environment. We consider two-level hierarchy: at the first level several independent e-brokers allocate computational tasks to clusters according some criteria. At the second level each cluster schedules the parallel tasks assigned to it using heuristics, in particular based on strip-packing algorithms. The efficiency of proposed hierarchical managing is estimated by modeling and computational experiments. [1] Foster, C. Kesselman, editors. The Grid: Blueprint
for a future computing infrastructure, Morgan Kaufmann, San
Fransisco, 1999. |