Abstract:
One of the best things about research is when you discover that something you were sure was true is, in fact, false. And it is even better when you can prove it!

My first talk will discuss a stream of research I began as a graduate student, when I discovered that the behavior of multi-server queues facing heavy-tailed service time distributions is in fact very different - actually much better - than what I (and everyone else) thought. Specifically, we will explore how the stationary delay in multi-server queues (specifically FIFO GI/GI/k queues) depends on the distribution of the arriving jobs and the traffic intensity (average rate at which work arrives) at the queue. We will also discuss the intuition behind this behavior - why, in retrospect, this is the only logical way such queues could behave. We will then discuss related results which may include (time permitting): How the workloads of other components of the multi-server workload behave, what happens when you have multi-server queues with heavy tails in series, what this might mean for related single-server queues with dependent service times (GI/G/1 queues), and new (partial) results on how the delay in queues with integral load behave. We will conclude with some open problems which I think present attractive future research opportunities.

This talk is composed of results from joint work with Karl Sigman, Rein Vesilo, and Michele Dufalla.