MAC intelligence for adaptive multimedia in 802.11 networks

Services in wireless networks must be capable of receiving information about the network and adaptively tune their transport parameters to the underlying networking conditions and technologies. A central problem in wireless transports is obtaining reliable metrics of congestion. How can lower layers assist transports and what is the performance tradeoff with pure peer-to-peer end-to-end solutions? We design and evaluate a lower layer assistance architecture. We focus on adding minimal intelligence to lower layers, according to the end-to-end principle. We find that we can adequately solve the measurement problem by minimal medium access control (MAC) assistance and describe an architecture that can aid transports over wireless links. The MAC-assisted solution is scalable to large number of flows, where the performance of pure end-to-end transports deteriorates rapidly. An improvement factor of 30-50 is exhibited in our experiments. We argue that including this minimal additional functionality in the MAC is sufficient for transports. It is also necessary when compared with end-to-end techniques.