Wireless, multi-hoping sensor networks with more than tens of motes often encounter traffic congestion. The occurrence of congestion subsequently results in loss of control over the bandwidth each mote effectively receives. In this paper we present a bandwidth management mechanism for sensor networks called Argus that incorporates congestion control and the ability to divide available bandwidth amongst traffic flows and as well as handle different traffic patterns. Argus focuses primarily on the transfer of data packets, and is implemented in the routing and transport layers. Since it independent of the data-link layer, it can therefore be applied to a network with different MAC protocols. Argus implements Extended Epoch-based Proportional Selection (EEPS), a distributed variant of Weighted Fair Queuing. EEPS performs scheduling of per-child queues, with the associated weights primarily derived from the flow sizes of downstream motes. Since a mote only requires knowledge of each child mote's total flow and not each individual flow's size, and since queues need be maintained on a per-child rather than a per-flow basis, Argus is scalable and is ideal for resource-constrained sensor networks. We demonstrate that Argus is able to achieve its goals of bandwidth management using a 26-mote mica2dot testbed.




Download Full History