Capacity of multi-channel wireless networks

This paper studies how the capacity of a static multi-channel network scales as the number of nodes, n, increases. Gupta and Kumar have determined the capacity of single-channel networks, and those bounds are applicable to multi-channel networks as well, provided each node in the network has a dedicated interface per channel.In this work, we establish the capacity of general multi-channel networks wherein the number of interfaces, m, may be smaller than the number of channels, c. We show that the capacity of multi-channel networks exhibits different bounds that are dependent on the ratio between c and m. When the number of interfaces per node is smaller than the number of channels, there is a degradation in the network capacity in many scenarios. However, one important exception is a random network with up to O(log n) channels, wherein the network capacity remains at the Gupta and Kumar bound of Θ(W√noverlog n) bits/sec, independent of the number of interfaces available at each node. Since in many practical networks, number of channels available is small (e.g., IEEE 802.11 networks), this bound is of practical interest. This implies that it may be possible to build capacity-optimal multi-channel networks with as few as one interface per node. We also extend our model to consider the impact of interface switching delay, and show that in a random network with up to O(log n) channels, switching delay may not affect capacity if multiple interfaces are used.