Dynamic channel-aware bandwidth management in IEEE 802.11 networks

One of the major requirements to support multimedia data in a dynamic environment such as a IEEE 802.11 wireless network is to provide the application minimum throughput guarantees. In this work, we present a cross-layer architecture that enables the provision of rate guarantees to multimedia applications, while still catering to best-effort traffic. We perform centralized arbitration of the shared bandwidth resource, and scale this centralized architecture to different 802.11 topologies. The major components of our architecture are a channel quality monitor, a central bandwidth manager (BM) that implements a channel allocation policy to distribute channel time to competing flows, and a rate-control mechanism that adjusts the traffic being injected into the network by a flow in accordance with its allocated fraction of the channel. The channel quality monitor is co-located with the IEEE 802.11 MAC protocol, but does not alter it in any way. It detects changes in medium contention and in channel fading errors as perceived by each network flow, since these phenomena are both time- and location-dependent. We assume adaptive multimedia applications that can function at multiple qualities. The bandwidth requirements of the application are translated into channel time requirements, taking into account the quality of the channel. The channel time requirement indicates the fraction of unit time the application needs to actively transmit its data in order to satisfy its throughput requirements. We present several policies for channel allocation by the central BM, in this work: fair, price-based, and utility-based. All our channel allocation policies endeavor to provide minimum throughput guarantees to the network flows. In the absence of wireless fair scheduling at the MAC-layer in current IEEE 802.11 products, we propose leaky bucket rate-control at the higher layers of the OSI protocol stack to control the quantity of traffic entering the network. We perform simulation and testbed experiments to evaluate the performance and demonstrate the feasibility of our overall architecture. Our experiments show that we incur low protocol overhead in providing statistical throughput guarantees to all network flows. Moreover, we find that our basic centralized architecture is flexible enough to work more-or-less unchanged with different IEEE 802.11 wireless network topologies.