Fast scheduling for optical flow switching

Abstract

Optical Flow Switching (OFS) is a promising architecture to provide end users with large transactions with cost-effective direct access to core network bandwidth. For very dynamic sessions that are bursty and only last a short time (~1S), the network management and control effort can be substantial, even unimplementable, if fast service of the order of one round trip time is needed. In this paper, we propose a fast scheduling algorithm that enables OFS to set up end-to-end connections for users with urgent large transactions with a delay of slightly more than one round-trip time. This fast setup of connections is achieved by probing independent paths between source and destination, with information about network regions periodically updated in the form of entropy. We use a modified Bellman-Ford algorithm to select the route with the least blocking probability. By grouping details of network states into an average entropy, we can greatly reduce the amount of network state information gathered and disseminated, and thus reduce the network management and control burden to a manageable amount; we can also avoid having to make detailed assumptions about the statistical model of the traffic.