A Non-Cooperative Game Approach For Availability-Aware Task Scheduling In Cloud Computing

Cloud computing task scheduling algorithm affects the benefits of cloud service providers and users.Efficient task scheduling algorithm improves the overall benefit of cloud computing system.In task scheduling,the overload of servers increases the queuing time of tasks,reducing the benefits of the cloud computing system.Therefore,reducing load imbalance is one of the goals of efficient task scheduling algorithms.This thesis proposes a availability-aware Task Scheduling Algorithm(AATSA)for cloud computing.In this Algorithm,users can reduce the load imbalance of the system by sensing the Availability of virtual machines.Specifically,In this thesis,the task scheduling process is regarded as a user game process in a competitive relationship,so that the task scheduling scheme can be obtained with the help of game theory.Users play with other users to minimize their task response time and task usage cost.In the process of user games,the processing rate of virtual machines is considered by users to decrease linearly with the increase of the number of tasks.Changes in the processing rate of virtual machines lead to changes in response time and usage cost,and promote more even task allocation.The proposed AATSA algorithm is simulated on the Cloud Sim platform,and compared comprehensively with Min RT,Min Cost and CALBA task scheduling algorithms based on non-cooperative game in three aspects of system task arrival rate,system scale and system heterogeneity.Especially,when the task arrival rate is 0.8,AATSA reduces4.69% and 8.99% of total response time compared with Min RT and Min Cost,but increases4.22% of total response time compared with CALBA.AATSA reduces 6.22% and 6.49%of total response time compared with Min RT and Min Cost,but increases 4.20% of total usage cost compared with CALBA.AATSA reduced the load imbalance by 51.37%compared with CALBA.AATSA also increases total response time and total cost of use and reduces system load imbalance under different system sizes and heterogeneity.In conclusion,the AATSA algorithm proposed in this thesis can effectively reduce the load imbalance of any system task arrival rate,any system size and any system heterogeneity degree under the condition of increasing the total response time and total usage cost.