Load Balancing in Cloud Computing Environment by Considering the Dependency among Tasks and Using Adaptive Genetic Algorithm

The increasing computing needs leads to an increase in the importance of using cloud computing day by day. Cloud computing is a computing model based on computer networks that presents a new pattern to supply resources, so that users request or release resources based on their needs. When the requests for computing resources increase, proper distribution of resources becomes very important, because if a computational unit has a large number of tasks and the other one is almost idle, resources are not used well and also makespan would be very high. Therefore, in order to overcome this problem, load balancing technique is used. In general, from the computing point of view, the process of distributing the loads on the processing units in a balanced way is called load balancing. In most researches, interactions between running tasks are not considered, so if the tasks in interaction with each other are located in separate processing units in a distributed network, the interactions between them would be effective on makespan. The aim of this research is to present an approach which can achieve a desirable load balancing in the network, such that the makespan and idle time of machines minimized, taking into account the interactions between the tasks. For this purpose, the genetic algorithm is used. The obtained experimental results show that localizing the interactions will have a significant impact in reducing the makespan.