Abstract
We present a study of the effects of active queue management (AQM) on the average queue size in routers. In this work, three prominent AQM schemes are considered: packet classification, checking service level agreements (SLA) and queue scheduling. This paper presents several adaptive resource sharing models that use a revenue criterion to allocate bandwidth in an optimal way. The models ensure QoS requirements of data flows and, at the same time, maximize the total revenue by adjusting parameters of the underlying schedulers. Deficit round robin (DRR) and deficit weighted round robin (DWRR) scheduling techniques have shown their ability in providing fair and weighted sharing of network resources for network devices. However, they are unable to use the total allocated network bandwidth even in burst traffic. In this paper, we propose a negative-deficit weighted round robin (N-DWRR) technique as a new packet scheduling discipline to improve the bandwidth utilization rate without increasing the total latency. A fully hardware packet scheduler has been implemented and verified as part of an intellectual property core. This is motivated by the fact that the design and analysis of hardware/software architectures for such techniques requires new models and methods, which do not fall under the domain of traditional embedded-systems design.