A library for load balancing in Master/Slave applications on a LAN-WLAN environment

Abstract

A combination of a Local Area Network (LAN) and a Wireless LAN (WLAN) is a useful environment for doing parallel computing. We use this infrastructure to implement Master/Slave parallel distributed programs which exhibit strict data dependences among iterations and parallel calculation inside an iteration. In this environment it is necessary to consider the unbalance both processing power and communication bandwidth of fixed and portable computers to avoid long idle times. To solve this problem, in our previous work we developed an efficient load balancing method that considers the heterogeneity both in computing and in communication, and besides, it keeps in mind that the number of processes can vary at run time (processes spawned on portable computers). To program this method is a hard task because it implements a complex mechanism to estimate the data distribution which uses a framework based on Simple Network Management Protocol (SNMP) that has to synchronize with the parallel program to share performance information about computers. For that reason, in this paper we present a library that implements our method of load balancing to ease the programming.