Associated Task Scheduling Based on Dynamic Finish Time Prediction for Cloud Computing
Yuqi Fan, Liping Tao, Jie Chen · 2019
Cloud computing has emerged as an increasingly indispensable and highly demanded platform for various applications, as cloud computing allows for on demand resource provisioning and allocation. The associated tasks composing a job processed by cloud computing need to be executed under ordering constraints for correctness or consistency. The associated tasks are executed on different servers and communication is required to transfer the data between the servers, while the processing capacity of and the communication capacity between different components underlying the cloud computing platform may show great heterogeneity. Therefore, efficient scheduling for the associated tasks is critical for achieving high performance in cloud computing systems. In this paper, we tackle the problem of associated task scheduling for cloud computing with the aim to minimize the makespan of the job, when excessive diversities are present in the computing and communication components. We propose a Dynamic Priority List Scheduling (DPLS) algorithm based on dynamic task finish time prediction. The algorithm dynamically predicts the remaining execution time for each task to be scheduled according to the server allocation of the previously scheduled tasks, and decides the next task to be scheduled and the server allocation based on the previously task scheduling result. We conduct experiments through simulations on randomly generated associated tasks and real-world applications. Experimental results demonstrate that the proposed algorithm is promising.