May 172012

This is one of the variety of the project I’m currently into.

Currently deployment of translucent optical networks is the most promising short term solution in optical backbone network to reduce cost and energy consumption within the network. But at the same time some problems could appear, like inflexibility and coarse granularity in the Wavelength Switched Optical Network (WSON). So how to deal with it – apply Optical Burst Switching (OBS), or even better ( as not all modification of this switching could be good enough) Translucent OBS (T-OBS)

What is actually T-OBS?  

It’s an architecture in which core nodes switch incoming data burst to their output ports in an all-optical fashion or through regenerators when signal regeneration is required. It is. in some kind of, a bridge between transparent and opaque solutions of OBS.

The background information for the project is taken from the Oscar Pedrola Paper.

Project itself

In such kind of network with OBS switching technology, necessity of solving RRPD problem could apear.

What is RRPD?

RRPD is a Routing and Regenerators Placement and Dimensioning problem made out of two parts. The first part of this problem is the minimization of average paths that goes through one link. In the second part it is searched where to place the necessary regenerators in order to reduce the number of them in the network and provide an acceptable signal-to-noise ratio.

As the problem is very complicated and needs a lot of resources, it is a  good idea t split the work:

  • starting from the MILP for the routing problem, which was already implemented in CPLEX (code is available on the wiki). As the solution of this problem suppose to be precise, in the paper it was solved only through ILP. In this project we went a bit far and suggest an heuristic for this part as well. The main idea is to compare both ILP and heuristic, and perform an analysis based on it.
  • the next part of the project is to deal with regenerator placement and dimensioning. After a few hours of researching on the most efficient heuristics for this problem – Biased Random Key General Algorithm (BRKGA) was chosen. Even though this problem in the paper was solved only by mean of heuristic, we are planning to implement ILP for this problem in CPLEX and compare the result and performance of these algorithms.

All further details on the project could be found on my wiki.