Implementation, evaluation and application of distributed heuristics for optimizing SLA admission
Date
2003
Authors
Shelford, Steven John Roy
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Reservation of bandwidth is necessary to absolutely guarantee the quality of service (QoS) of multimedia sessions in interconnected networks. We explore the use of the Multidimensional Multiple-choice Multi-knapsack (MMMKP) problem to solve the problem of optimal admission and adaptation control of service level agreements (SLAs) with multiple levels of QoS. The goal of admission and adaptation is to maximize a function of system utility, subject to constraints on resource allocation (link bandwidth) imposed by the need to respect all QoS guarantees. We first propose P-HEU, a new distributed heuristic designed with a parallel algorithm approach. Next, to evaluate several heuristics, we outline the implementation of a simulated distributed optimal SLA admission and adaptation controller, dSLAOpt. Upon evaluating the heuristics, we propose a modification of a previous heuristic, MF A-HEU, which is fast, and yields high system utility. The thesis concludes with the introduction of a unified admission model for the reservation of both network and multimedia server resources for multimedia sessions.