Load balancing scheduling algorithms for virtual computing laboratories in a Desktop-As-A-Service Cloud Computing Services

The present research work is interested in the study of the virtual computing laboratories in a Desktop-As-A-Service (DAAS) scheduling in a cloud computing environment. The issue emanates from a real-world study in which the Saudi Electronic University expresses a need to efficiently schedule the labs in predefined sessions for one of its colleges over all its branches. The problem under investigation consists, on the one hand, in load balancing the assignment of the labs to the sessions regarding the number of participants, and on the other hand, in determining the types and number of hosts required to ensure the best sessions hosts assignments. While the load balancing is evaluated by measuring the distance between the maximum and the minimum number of the virtual machines load, the maximum number of the virtual machines load, and the variance of number of the virtual machines load, the efficiency of the host selection and assignments are assessed in terms of the unbalance of the remaining capacity of each used host. We propose linearized mathematical models for the subproblems and the whole problem with the different objective functions. We firstly suggest a two-stage optimization approach in which the mathematical models via CPLEX are solved. Then, we deal with the combined method which does not divide the whole problem into its two sub-problems. Next, we run the combined model on CPLEX, and we propose two heuristics to have a feasible good solution in reasonable time. Subsequently, we compare the obtained results for diverse sizes of instances we generate according to the University features. The proposed heuristics are therefore proven to be a compelling solution to the powerful software CPLEX's failure to solve the problem for large-scale cases. Moreover, the obtained results demonstrate that they outperform those currently in use in the University.

Automated summary

This research focuses on the study of virtual computing laboratories in Desktop-As-A-Service (DAAS) scheduling in a cloud computing environment. The objective is to efficiently schedule the labs in predefined sessions, considering load balancing and host selection. Mathematical models and optimization approaches are proposed, and heuristics are developed to solve large-scale cases. The results demonstrate the compelling performance of the proposed heuristics.