Dynamic Multiservice Load Balancing in Cloud-Based Multimedia System

Consider a centralized hierarchical cloud-based multimedia system (CMS) consisting of a resource manager, cluster heads, and server clusters, in which the resource manager assigns clients´ requests for multimedia service tasks to server clusters according to the task characteristics, and then each cluster head distributes the assigned task to the servers within its server cluster. For such a complicated CMS, however, it is a research challenge to design an effective load balancing algorithm that spreads the multimedia service task load on servers with the minimal cost for transmitting multimedia data between server clusters and clients, while the maximal load limit of each server cluster is not violated. Unlike previous work, this paper takes into account a more practical dynamic multiservice scenario in which each server cluster only handles a specific type of multimedia task, and each client requests a different type of multimedia service at a different time. Such a scenario can be modelled as an integer linear programming problem, which is computationally intractable in general. As a consequence, this paper further solves the problem by an efficient genetic algorithm with an immigrant scheme, which has been shown to be suitable for dynamic problems. Simulation results demonstrate that the proposed genetic algorithm can efficiently cope with dynamic multiservice load balancing in CMS.