engleski

Load Balancing: Moving Toward Mobility and Intelligence

Development of effective techniques for task distribution on multiple processors or nodes is considered as one of the biggest issues in parallel and distributed operating environments. In a distributed system, statistical fluctuations in arrival and service patterns across various nodes cause imbalances in load, where one or more nodes may be operating much below capacity while others may simultaneously be overloaded. Thus, an improvement in the overall system performance can be achieved by the process of moving jobs from overloaded to underloaded nodes, which is called load balancing. Over the last several years, a number of load balancing methods have been proposed. This article describes the generic model of the load balancing system where load balancing methods have features of autonomous stationary agents. Furthermore, simulation of a distributed system with load balancing capabilities, which has the features of a multi-agent system, is explained and results from different methods&#8217 ; performance comparison are commented. The majority of invented methods have been using client/server technology. Although the software agents paradigm has appeared recently, it has a significant influence on the world of distributed processing. This article also gives a short overview of the achieved results in agents-based load balancing. Furthermore, real-time simulation of an agent load balancing system is described and results from different methods&#8217 ; performance comparison are commented.

load balancing; software agents; multi-agent system

nije evidentirano