A learning analytics approach for job scheduling on cloud servers

Abstract

Learning analytics improves the teaching and learning procedures by using the educational data. It uses analysis tools to carry out the statistical evaluation of rich data and the pattern recognition within data. This chapter, firstly, describes four learning analytics methods in educational institutions. Secondly, it proposes a learning analytics approach for job scheduling on cloud servers, called LAJOS. This approach applies a learning-based mechanism to prioritise users’ jobs on scheduling queues. It uses the three basic attributes “importance level”, “waiting time” and “deadline time” of various jobs on cloud servers. The cloud broker acts as a teacher and local schedulers of cloud sites act as students. The broker learns to local schedulers how to prioritise users’ jobs according to the values of their attributes. In the deployment phase, the effect of the above attributes on the system throughput is studied separately to select the best attribute. In the service phase, users’ jobs are prioritised by computer systems according to the selected attribute. Simulation results show that the LAJOS approach is more efficient compared to some of the job scheduling methods in terms of schedule length and system throughput.