An Improved Task Scheduling Algorithm Based on Potential Games in Cloud Computing

The task scheduling problem of Cloud computing was modeled as an potential game. The mapping problem of users' tasks to virtual machines was abstracted into a path selection problem in traffic network. Each task agent was viewed as a selfish participant, who competed with each other for virtual machine. Considering the virtual machines' capacity, we proves the game is a potential game which exists at least one Nash equilibrium. The consistency of Nash equilibrium and the minimum value of potential function is proved. Finally, two improved task scheduling algorithm based on potential game are proposed. By our algorithms, the system can reach to a stable state eventually. Meanwhile, the algorithms can guarantee that the system load balancing level is adaptive to the amount of users' task changing.