Understanding the maximum weight policy
(joint work with Damon Wischik, UCL.)
The maximum weight policy schedules queues as per their weight -- a
function of queue-size -- in a constrained packet network. In recent
years, it has been quite popular due to wide applicability and
excellent performance despite its myopic nature.
In this talk, we will be interested in understanding the effect of the
choice of weight (in terms of the function of queue-size) on the
network performance. We will consider a parameterized class of weight
function, f(x) = x^a, for a > 0.
We shall discuss the monotonicity of performance of this class of
algorithms with respect to a. Next, we will discuss the optimal
algorithm, based on logarithmic weight function. Finally, we will
discuss its implications in the context of flow-level model of the
bandwidth sharing network. Our results are based on critical fluid
model of the network.