THEORETICAL FOUNDATIONS OF CREATING SELF-ORGANIZING DISPATCHERS OF DISTRIBUTED SYSTEMS BASED ON A MULTI-AGENT SOCIO-INSPIRATIONAL APPROACH

Abstract

This article describes new principles of organization, methods and algorithms for the functioning of the Distributed System (DS) dispatcher, which allow allocating and reallocating resources with dynamically changing parameters between incoming tasks in order to minimize their execution time. The main problem that does not allow today to effectively estimate the execution time of tasks in a heterogeneous DS directly follows from the distribution of the system: each of its elements has partial independence and may differ significantly from others, moreover, in the process of operation, its capabilities may change, and all this is essential. affects the efficiency of distribution of tasks between DS nodes and the time it takes to complete tasks. The article proposes a new approach to organizing a DS dispatcher, based on the application of the theory of multiagent systems and socio-inspirational (based on accepted in human society) methods: DS users place their tasks on special nodes – bulletin boards, a proactive software agent is placed on each DS node, which implements constant monitoring of the parameters of your site and search on message boards suitable for solving problems. At the same time, the agents participating in the solution of the common task form communities in which they plan the process of solving the task and the distribution of parts of the tasks to minimize the delay time for their solution. As a criterion for the effectiveness of the DS, it was decided to take the value of the average delay in the execution of functional tasks relative to the required points in time, respectively, the agents distribute tasks in such a way as to minimize the value of the specified criterion. This article includes an introduction, a formal statement of the task of scheduling DS resources, a review of existing approaches to organizing a DS dispatcher, a description of the proposed multi-agent solution to the task of scheduling DS resources using a socio-inspirational approach, an algorithm for the operation of a distributed system and its elements, a description of the application of a socio-inspirational approach in relation to the task scheduling process. and conclusion. The main advantages of the proposed approach include: the ability to use reliable and up-to-date information about the specialization and current performance of resources in dispatching; high fault tolerance due to the absence of DS elements, failure of which leads to a complete loss of DS performance; the possibility of flexible scaling of the DS (increasing the number of resources), achieved by decentralizing the dispatching process.