The present invention provides a method, computer-readable medium and device for dynamically managing allocation of bandwidth in a packet network using a Dynamic Setting Scheme (DSS) for Class Based Queuing (CBQ). The method includes the steps of maintaining a minimized reserved portion of bandwidth to minimize delay jitter and maximizing a shared portion of bandwidth to maximize overall bandwidth utilization. Allocation of bandwidth may be based on a sharing tree hierarchical scheme that provides for temporary borrowing of bandwidth by real-time applications from bandwidth of non-real-time applications and blocks borrowing of bandwidth by non-real-time applications from bandwidth of real-time applications. DSS typically provides for using measurable parameters, such as queue length and number of borrowing attempts per a predetermined length of time/observation window, as control triggers for implementing adjustment of bandwidth allocation. In one embodiment, the steps include measuring a predetermined parameter at predetermined observation window times and dynamically adjusting allocated bandwidth for parent classes of real-time traffic by adjusting an average of the predetermined parameter to have a value within a predetermined stable region.
Methods, apparatuses and systems allowing for dynamic bandwidth management schemes responsive to the state of a plurality of access links in redundant network topologies. In one embodiment, the present invention provides a bandwidth management device that periodically queries routing systems associated with access links, conceptually grouped into a virtual access link, to monitor that load of the access links and, depending on the detected load, adjust the configuration of the bandwidth management device to avoid overloading one or more of the access links. Embodiments of the present invention increases network efficiency and help network traffic to flow more smoothly with higher throughput. In one embodiment, the dynamic link control functionality is invoked when any given access link reaches a threshold capacity level. Assuming that network traffic will scale in the same ratio as presently observed, the present invention calculates the maximum traffic that can be let through so that no network interface or access link is overloaded.
Systems and methods for two-rate scheduling over a class hierarchy wherein controlled rates are delivered in a smooth manner, even over short time intervals. A minimum rate scheduler and an excess rate scheduler are employed. The minimum rate scheduler and/or the excess rate scheduler, employ special binary search trees to make selections at each parent node in the class hierarchy.
