Pull vs. Hybrid: Comparing Scheduling Algorithms for Asymmetric Time-Constrained Environments.
Jesús Fernández-Conde, Daniel Mozos · International Conference on Wireless Networks · 2008
Abstract Data broadcasting has been recognized as a very effectivedata delivery mechanism in mobile computing environmentswith an unlimited number of users and changing demandpatterns. A fundamental server design issue is thescheduling algorithm needed for the construction of thebroadcast program. In most situations, users requireinformation before a certain time, and therefore the serverscheduling algorithm should consider the existence oftiming constraints. In this paper we analyse the theoreticallower bound for deadlines missed in strong server overloadconditions. Considering the results of the theoreticalanalysis, this paper compares two recent schedulingalgorithms, pull-based SINα and hybrid (push-pull)AHB, both proved to perform efficiently in asymmetrictime-constrained communication environments in whichthe user access distribution is dynamic. Performance isevaluated by simulation. 1 Introduction Recent technical advances in the development of portablecomputing devices and the rapidly expanding wirelesstechnologies have made mobile computing a reality. Thedesign of information servers with capabilities to providedifferent types of information (e.g., traffic reports, stoc kquotes, weather, news, airline schedules) to a potentiallyunlimited number of mobile users carrying portable devicesgives rise to significant challenges. In these emerginginformation-access applications, there exists an asymmetryin the communication between server and clients, mainlydue to the fact that the available bandwidth for transmissionand the flow of information from the server to the clients(downlink direction) are much larger than the ones fromclients to server (uplink direction).In most situations, mobile users require the desiredinformation by a certain time, and therefore the servershould consider the existence of timing constraintsassociated with the requests of the data. For example, let