Title :
Analysis of cluster load balancing using simulation
Author :
Turcinhodzic, Razija
Author_Institution :
Dept. of Comput. Sci. & Inf., Univ. of Sarajevo, Sarajevo, Bosnia-Herzegovina
Abstract :
The process of load balancing redistributes the load among the processors (servers). It improves the performances of parallel and distributed systems. The increasing demand for computing power has made this area interesting. This paper presents the analysis of three types of cluster load balancing: cyclic, random and the minimal queue method. The analysis is performed using simulation. This paper shows that the distribution of arrival intervals and the service time distribution significantly change the performance of a cluster. However, apart from that, load balancing using the minimal queue method always achieves the best performances in comparison with the other two methods mentioned in this paper. All load balancers do not use the minimal queue method when opting for one out of these three methods. The reasons for this are specific demands that servers have to meet.
Keywords :
file servers; microcomputers; queueing theory; resource allocation; arrival intervals distribution; cluster load balancing; computing power; distributed systems; load balancers; minimal queue method; parallel systems; processors; servers; service time distribution; Algorithm design and analysis; Analytical models; Computational modeling; Load management; Load modeling; Queueing analysis; Servers; cluster; load balancing; queues system model; simulation;
Conference_Titel :
Telecommunications and Signal Processing (TSP), 2013 36th International Conference on
Conference_Location :
Rome
Print_ISBN :
978-1-4799-0402-0
DOI :
10.1109/TSP.2013.6613936