Optimizing Concurrency Performance of Complex Services in Mobile Environment

Abstract

Hosting services on mobile devices has been considered as the key solution for domains that have special requirement on portability, timeliness, and flexibility on service deployment. Typical examples include, among many others, military, music, healthcare, gaming, and data sharing. Despite the recent boom of mobile computing makes service deployment in mobile environment possible, significant challenges arise due to the limitations in existing mobile hardware/software capable of managing resource intensive applications. The situation gets worse when managing complex services that allow concurrent clients and requests. This paper addresses the issue related specifically to concurrency control improvement in mobile web servers to support the mobile deployment of complex services. The authors identify key factors that affect a system to respond a request, including request related factors, system resource related factors, and context. Based on this, the authors propose a dynamic heavy request classification model (DHRC) to estimate the heaviness of an incoming request using machine-learning methods. The heavy request will be detected, which requires relatively heavy system resources of the mobile server to generate a response. The authors design a dynamic request management strategy (DRMS), which reduces the number of discarded requests by adding heavy requests to a queue and processing them asynchronously. The proposed solution is implemented on Android-based mobile devices as an extension of I-Jetty web server. Experimental studies are conducted and the result indicates the effectiveness of our solution.