|
Student: Wei Liu, M.Eng Student
Supervisor: Prof. Mark Coates
Abstract: Click here
Research problem:
- Wide-area network has substantial propagation delays.
- In an AAPN, control functionality will be implemented primarily at the edge of the photonic core.
- The distributed edge nodes cannot measure the performance of the global network.
- The core node is performing time slot switching.
- Requests for additional time slots must be initiated by the edge nodes before the demand arises
- Need to perform short-range prediction of an upper bound on source-destination traffic demands.
- Request and relinquish slots based on predicted upper bound.
Traffic Modelling Approach:
- Intra-AAPN Modelling time scales are shorter than network round-trip-times so traffic models can ignore long-range dependent effects
- A Markov-modulated Poisson model for total outgoing traffic from an edge node
- A Gravity model for the splitting of a source i traffic between individual destinations.
- The parameter in the gravity model is updated using feedback from a monitor at the core.
- Modelling approach allows us to capture both temporal and spatial information.
|