JUSTIFICATION FOR CHOOSING A DISCIPLINE FOR SERVICING REQUESTS IN DISTRIBUTED INFORMATION PROCESSING SYSTEMS

Abstract

The article examines approaches and the most common application service disciplines (algorithms for application selection for allocating time on the processor. The classification of scheduling algorithms is given. Priority and non-priority disciplines are distinguished. Service algorithms with both one queue and several queues are presented. The advantages and disadvantages of the presented disciplines, as well as the peculiarities of their application, are shown. From the presented analysis, it can be concluded that all the described algorithms perform planning in time, as they are focused on a computing environment with one processor.However, it is necessary to take into account that modern computing systems of information processing have characteristics inherent in distributed systems. The peculiarity of modern systems is that the execution environment (computing resources) can provide many resources simultaneously for many applications. In them, both computing resources and tasks are distributed in space. Such features are characteristic of global GRID and CLOUD information processing systems. In addition, tasks and resources can have characteristics of homogeneity and heterogeneity. With such characteristics, the disciplines intended for single-processor systems cannot be used for distributed information processing systems. In such systems, the scheduler has information about the number of requests requiring service, as in known service systems, and information about the number of resources that have the ability to service the requests. Before this, the scheduler has information about the requirements of each application for the intended resource. In addition, the scheduler has information about the characteristics of each resource. Such information about requests and resources can allow the scheduler to provide optimal re-allocation of requests for resources, taking into account the specifics of the request requirements and resource characteristics. Thus, a new approach to the design of new types of planners is needed in a distributed information processing system – that is. spatial planners. The article presents a theoretical rationale for a new approach to the selection and practice of a spatial planner. The article shows that application-resource relations in distributed systems can be represented in the form of a bipartite graph. And the task of distributing the request for a resource is reduced to the task of finding the maximum pairing in a weighted or unweighted bipartite graph.