22 March - 26 March 2010
Isaac Newton Institute for Mathematical Sciences, Cambridge, UK
Workshop Organisers: Takis Konstantopoulos (Heriot-Watt) and Kavita Ramanan (Brown University).
Bruce Hajek (University of Illinois), Frank Kelly (Cambridge), P.R. Kumar (University of Illinois), Tom Kurtz (University of Wisconsin-Madison), Jean Mairesse (LIAFA), Philippe Robert (INRIA), John Tsitsiklis (MIT) and Ruth Williams (University of California, San Diego)
in association with the Newton Institute programmeStochastic Processes in Communication Sciences (11 January to 2 July 2010) Participants | Application | Accommodation and Cost
Title: Self-adaptive congestion control for multi-class intermittent transmissions in a network.
This study extends a previous Markovian model for permanent transmissions emitting through a network. Numerous connections (or users, customers, ...) are divided into classes by attributes such as routes, traffic characteristics, quality of service requirements, etc. An inactive connection becomes active when it acquires data to transmit, and becomes inactive again after having completed the transmission. Active connections create congestion in the network, and control it in a delocalized way, by each adjusting its output to the resulting delays and losses it incurs, using an Additive Increase, Multiplicative Decrease (AIMD) algorithm similar to TCP.