Network Model of It-Infrastructure Resource Management

Abstract

A formal model of a system for planning tasks in an IT infrastructure based on temporary Petri nets is proposed. The proposed model consists of a finite set of executors, conventionally called servers, a finite set of operations that can be performed by servers and two reflections. The servers are divided into input servers, where initial data are generated, production servers, where tasks are performed, and output servers, where the results of job execution by working servers are stored. The first of the reflections determines a subset of the operations for each production server, that this server can perform, and the second reflection determines the time of each operation’s execution on the corresponding server. Since it is assumed that the same operation can be performed on different servers, the execution time of such an operation by different servers can also be different. The main problem that is being solved is the optimal planning of operations in such a system. The proposed formalism is illustrated by the example of a system with two input servers, three production servers and two output servers with the interpretation of the search for the optimal path by the genetic algorithm in a directed graph with the calculation of the deviation for two tasks. The first task performs search for the graph using two optimality criteria, and the second task searches for the same graph with the other two optimality criteria. The choice of the optimal execution plan is modeled by a temporary Petri net, and with the help of its analysis the best execution plan for the tasks by time characteristics is found.