This letter proposes a new weighted fair queueing algorithm, which adjusts dynamically each flow's service probability according to its weight and average packet length and then uses the service probability parameters to implement fair queueing. This solves the main drawback of traditional weighted fair queueing algorithms--the packet-based tracing of weight parameters. In addition, this letter proposes a novel service probability calculation method which solves the unfairness problem induced by the variable packet length.

Supporting Quality of Service (QoS) over the Internet is a very important issue and many mechanism have been already devised or are under way towards achieving this goal. One of the most important approaches is the class-based architecture, which provides a scalable mechanism for QoS support in a TCP/IP network. Class-based service differentiation can be realized without resource reservation, admission control and traffic policing. However, the resulting services are only relative. While it is, in principle, not feasible to provision for absolute guarantees without admission control and/or traffic policing, such a service can be reasonably well emulated using adaptive rate allocation at the link scheduler of routers. In this paper, we propose mechanism for link scheduler of router that achieve emulated absolute and other relative guarantees using dynamic weighted fair queueing (DWFQ) combining with class packet dropping. The weights of DWFQ are frequently adjusted to current load conditions and based on prediction of realistic class traffic. These mechanisms can realize many approaches to QoS guarantees and class-based differentiation.

In the past, a number of scheduling algorithms that approximate GPS, such as WFQ, have been proposed and have received much attention. This class of algorithms provides per-flow QoS guarantees in terms of the bounded delay and minimum bandwidth guarantee. However, with O(log N) computational cost for each new arrival scheduling, where N is the number of backlogged flows, these algorithms are expensive to implement (e.g., in terms of scalability). Moreover, none of them addresses the issues of delay distribution and jitter. In this paper, we propose a new traffic scheduling discipline called Jitter Control Frame-based Queueing (JCFQ) that provides an upper bound for delay jitter in the case of rate-controlled connections, such as packet video streams and IP telephony, while guaranteeing bounded delay and worst-case fair weighted fairness, such as in the WF2Q algorithm, but with O(1) complexity in selecting the next packet to serve, assuming that the number of flows is fixed. Three different algorithms for slot or service order assignment between flows are proposed: Earliest Jitter Deadline First (EJDF), Rate Monotonic (RM) and Maximum Jitter First (MJF). In these algorithms, delay jitter is formulated into the virtual finish time calculation. We compare the fairness, delay and jitter performance of the JCFQ with that of the MJF algorithm with WF2Q via simulation. The results show that with proper choice of the slot size, JCFQ can achieve better flow isolation in delay distribution than can WF2Q.

This paper discusses two approaches to statistical multiplexing: rate envelope multiplexing, allowing resource sharing with small delays for low peak rate connections, and rate sharing, based on the use of large multiplexer buffers to ensure high link utilization for high speed data traffic. We argue that the weighted fair queueing scheduling algorithm provides an efficient means for combining both kinds of multiplexing in the B-ISDN. A feasible implementation known as Virtual Spacing is outlined. We illustrate the flexibility of the proposed scheme by showing how different service categories could be provided.