Queuing in a Random Environment
Tuesday, March 17, 1998
1:00-2:00 p.m.

We discuss a class of queueing network models with servers which are randomly modulated and can be dynamically allocated to various queues. As a basic model we consider a system of Q parallel queues and a single server which can be in one of N possible states at any given time. There are job flows into the queues. When the server is in a particular state n, queue q will receive service at rate r(n,q) if the server is allocated to it. The server state is randomly modulated by an extraneous agent driving the random environment. The core problem is how to allocate the server to the queues so as to maximize the system throughtput given the statistics of the server state modulation. We address this issue and a few other related ones. We then extend the model to networks of nodes of the type specified above. The problem then is how to manage the servers so as to maximize the job flow through the network. The motivation for studying such models comes mainly from wireless communication networks, where server modulation is induced by fluctuating interference.
UC Berkeley Networking
Ashwin Pananjady and Orhan Ocal
Last Modification Date: Wednesday, February 10, 2016