Attitudinal data based server job scheduling using genetic algorithms: Client-centric job scheduling for single threaded servers

Abstract

With the evolution of web applications towards real time setups, like web sockets, the need for more efficient and setup-specific scheduling techniques escalates. Thus, an effective task scheduling mechanism becomes the prime necessity for any server handling large number of requests concurrently. The existing task scheduling algorithms do not suffice this purpose due to their focus on being generic and minimizing the execution time, while failing to use the characteristics of the system architecture and job profiles with the associated user-agent utility. Additionally, the traditional setups also fail to exploit the knowledge derived from the job profiling, considering the fact they are designed to be more generic. In order to come up with a comprehensive and efficient mechanism, this paper focuses primarily on task scheduling for single threaded environments. Moreover, the proposed scheduling mechanism is driven by Genetic Algorithms (GA), while taking the profile(s) of requests to be served by the targeted server and then using the derived knowledge for enhancing the performance. The experimental results show that the performance of the target server (in terms of efficiency) improves, along with optimizing the client user-agent utility. Furthermore, the proposed model demonstrated significant improvement in user-agent utility while maintaining a viably constant time for computing the same under carefully designed test runs capable of being conclusive by providing enough coverage.