The convergence of communications, information, commerce and computing are creating a significant demand and opportunity for multimedia and multi-class communication services. In such environments, controlling the network behavior and guaranteeing the user's quality of service is required. A flexible hierarchical sorting architecture which can function either as a traffic shaper or a scheduler according to the requirement of the traffic load is presented to meet the requirement. The core structure can be implemented as a hierarchical traffic shaper which can support a large number of connections with a wide variety of rates and burstiness without the loss of the granularity in cells' conforming departure time. The hierarchical traffic shaper can implement the exact sorting scheme with a substantial reduced memory size by using two stages of timing queues, and with substantial reduction in complexity, without introducing any sorting inaccuracy.
By setting a suitable threshold to the length of the departure queue and using a lookahead algorithm, the core structure can be converted to a hierarchical rateadaptive scheduler. Based on the traffic load, it can work as an exact sorting traffic shaper or a Generic Cell Rate Algorithm (GCRA) scheduler. Such a rate-adaptive scheduler can reduce the Cell Transfer Delay and the Maximum Memory Occupancy greatly while keeping the fairness in the bandwidth assignment which is the inherent characteristic of GCRA. By introducing a best-effort queue to accommodate besteffort traffic, the hierarchical sorting architecture can be changed to a near workconserving scheduler. It assigns remaining bandwidth to the best-effort traffic so that it improves the utilization, of the outlink while it guarantees the quality of service requirements of those services which require quality of service guarantees. The inherent flexibility of the hierarchical sorting architecture combined with intelligent algorithms determines its multiple functions. Its implementation not only can manage buffer and bandwidth resources effectively, but also does not require no more than off-the-shelf hardware technology.
The correlation of the extra shaping delay and the rate of the connections is revealed, and an improved fair traffic shaping algorithm, Departure Event Driven plus Completing Service Time Resorting algorithm, is presented. The proposed algorithm introduces a resorting process into Departure Event Driven Traffic Shaping Algorithm to resolve the contention of multiple cells which are all eligible for transmission in the traffic shaper. By using the resorting process based on each connection's rate, better fairness and flexibility in the bandwidth assignment for connections with wide range of rates can be given.
A Dual Level Leaky Bucket Traffic Shaper(DLLBTS) architecture is proposed to be implemented at the edge nodes of Differentiated Services Networks in order to facilitate the quality of service management process. The proposed architecture can guarantee not only the class-based Service Level Agreement, but also the fair resource sharing among flows belonging to the same class. A simplified DLLBTS architecture is also given, which can achieve the goals of DLLBTS while maintain a very low implementation complexity so that it can be implemented with the current VLSI technology.
In summary, the shaping and scheduling algorithms in the high speed packet switches and DiffServ networks are studied, and the intelligent implementation schemes are proposed for them.
|
![[NJIT logo]](http://archives.njit.edu/vhlib/images/njit_logo.gif){kind=logo}
![[Van Houten Library logo]](images/vhl-top-image.jpg)
