| The deployment of digital satellite networks with integrated Internet Protocol (IP) support has emerged as an efficient mechanism for the delivery of multiplexed IP unicast and IP multicast transport. The multiplexing of such services, however, creates a need to resolve the inherently disparate rate and congestion control mechanisms of each transport type while satisfying the need to maintain high channel utilization with minimal packet loss. To satisfy these needs, a novel two-priority, class-based multiplexing scheme utilizing dynamic bandwidth allocation with prediction is developed. This scheme is coined Available Bit-Rate Multicast IP (ABRMIP) and is designed to provide dynamic bandwidth allocation based upon IP transport type. The predictability of network traffic will be established as the motivation for the development of the scheme's prediction mechanism, which is based on the used of wavelet adaptive filtering. Through online measurement of traffic flow, network packet transmissions are transformed to a discrete time series representation upon which the wavelet filter operates and control decisions are derived. By using online measurement combined with signal processing methodology, proactive flow and congestion control become possible and the problems inherent with reactive control operating on networks with large bandwidth-delay products, such as satellite, can be mitigated. Rather than reacting to congestion, ABRMIP with prediction identifies potential congestion in advance, allowing it to be avoided or reduced substantially. This is shown to improve system channel utilization while maintaining low levels of packet loss. |
