Issues in modelling optical networks

Stochastic Networks seminar, EE, Stanford University, 3 October 2001. [seminar] [slides pdf]

Abstract.

An all-optical network is likely to have two characteristics: high capacity, and limited flexibility. The high capacity comes about because optical switches are simple: they have no buffers and no high-speed per-packet classification, they just redirect beams of light. The limited flexibility comes about for the same reason; it has the consequence that neither packet-switched nor circuit-switched network models are entirely appropriate. In this talk I shall describe several possible models, which look at routing, multiplexing, and buffering.

The capacity of a burst-switched network

Rank Prize Funds mini-symposium on Terabit Optical Networks, Grasmere, 2-5 April 2001. [organizers] [slides pdf]

Abstract.

An all-optical network may not have sufficient wavelengths to be able to assign permanent connections between all pairs of end-nodes. A possible solution is to interleave data from different end-nodes on a single wavelength. There are various ways that this might be achieved, and they may be grouped under the general heading of burst-switching. It is natural to ask: what is the capacity of a burst-switched network? This talk describes some mathematical techniques that may be helpful in answering this question.