A variable local search based memetic algorithm for the load balancing problem in cloud computing

Load balancing (LB) is an important and challenging optimisation problem in cloud computing. LB involves assigning a set of services into a set of machines for which the goal is to optimise machine usages. This study presents a memetic algorithm (MA) for the LB problem. MA is a hybrid method that combines the strength of population based evolutionary algorithms with local search. However the effectiveness of MA mainly depends on the local search method chosen for MA. This is because local search methods perform differently for different instances and under different stages of search. In addition, invoking local search at every generation can be computationally expensive and compromise the exploration capacity of search. To address these issues, this study proposes a variable local search based MA in the context of LB problem. The proposed MA uses multiple local search mechanisms. Each one navigates a different area in search space using a different search mechanism which can leads to a different search path with distinct local optima. This will not only help the search to avoid being trap in a local optima point, but can also effectively deal with various landscape search characteristics and dynamic changes of the problem. In addition, a diversity indicator is adopted to control the local search processes to encourage solution diversity. Our MA method is evaluated on instances of the Google machine reassignment problem proposed for the ROADEF/EURO 2012 challenge. Compared with the state of the art methods, our method achieved the best performance on most of instances, showing the effectiveness of variable local search based MA for the Load Balancing problem.