A Heterogeneous Cluster Multi-resource Fair Scheduling Algorithm Based on Machine Learning

The resource scheduling of data center is a research hotspot of cloud computing. The exiting research work is concerned with the issue of fairness, resource utilization and energy efficiency, which are only applicable to the same cluster environment or specific application situations. First, the default scheduling algorithm (DRF) of Mesos is analyzed. The DRF algorithm does not consider machine performance and task types. Then, this paper presents a heterogeneous cluster multi-resource fair scheduling algorithm based on machine learning to solve the problem. The algorithm is to test the performance of the machine and use the machine learning method to classify the computing tasks and reach the goal of reasonable resource allocation. Finally, the experimental results show that the method presented in this paper not only ensures the fairness of resource allocation, but also makes the system more reasonable allocation of resources and further improves the system’s resource utilization.