Використання механізму активних об'єктів для паралельних обчислень в середовищі з невідчужуваними ресурсами

Abstract

The problem of using inalienable resource for GRID computing is discussed in this article. The analysis of approaches for planing tasks in parallel computing systems is presented: one and double-level manager, use the global and the local queue for the allocation of tasks, partitioning the nodes between the incoming jobs. The use of active objects is proposed for computing tasks in the context of the unreliability of transmission environment between the server and computing nodes. The article describes the system architecture, whch uses active objects, and explains the purpose of an active object's components.