| Demands for Quality of Service (QoS) on computers and networks are increasing with the emergence and rapid growth of multimedia network applications, e-commerce, and other transactions on the Internet. The current "best effort" service model for the Internet cannot provide guaranteed QoS. This study focuses on QoS provisioning, especially scheduling of computer and network requests or jobs to reduce their Waiting Time Variance (WTV) for stable QoS in timeliness. Stability in timeliness is the key issue in QoS because it helps to minimize variance of service time on a computer or network resource and thus enables dependable and predictable services. Most of the existing work on QoS focuses on the mean performance or worst-case performance of timeliness which does not lead to stable QoS. In this study, two job scheduling methods, Verified Spiral and Balanced Spiral, are developed to reduce WTV of jobs without priorities in a given batch on a single computer or network resource. The factors influencing WTV, such as the sum of processing times, distribution of job processing times, and scheduling methods, are investigated to assist the configuration of a computer or network resource for achieving a desired level of WTV. The weighted WTV problem, in which jobs have different weights or priorities, is also addressed. Two job scheduling methods, Weighted Verified Spiral and Weighted Simplified Spiral, are developed for the weighted WTV problems. WTV minimization can be applied to local computer and network resources, e.g. routers and web servers, and resources in other application fields where it is desirable to achieve QoS stability. |
