On the implementation of weighted fair queuing in high speed networks

Abstract

The fluid generalized processor sharing (GPS) algorithm has desirable properties for integrated services networks and many weighted fair queuing (WFQ) algorithms have been proposed to approximate the GPS. The implementation of WFQ algorithms using calendars (i.e. WFQ calendar) has been proved to be a practical way to realize the WFQ scheduler which supports a large number of queues with diverse rate requirements and at the same time maintains all the important properties of the GPS. However, a WFQ calendar will introduce errors due to the granularity of a calendar and the implementation methods and limitations. What the impacts of the errors are and how to design a WFQ calendar to meet the traffic management requirements become very important in the WFQ calendar design. But there have not been any discussion on this topic. In this paper, a WFQ calendar is modelled, and the implementation errors of WFQ calendars are discovered and analyzed. In addition, the impacts of the errors are discussed, and a few ways to reduce the errors are given.