Offline Routing and Regenerator Placement and Dimensioning for Translucent OBS Networks

By | May 17, 2012

Been a long time ūüėČ Seems I have been having a few days of rest (or the closest possible to it since I only managed to go to beach once!!).

Anyways, I’ve been doing a project that consists on implementing and evaluating the ILP and heuristics presented on paper “Offline Routing and Regenerator Placement and Dimensioning for Translucent OBS Networks“.

So what is a Translucent network?

…both the electronic switching networks and all-optical transparent networks have their own pros and cons. A translucent network just compromise these two types of networks to take the advantages of the two types of networks to achieve a good balance between cost and network performance. For cost saving, the translucent networks maximally pursue optical bypass opportunities when establishing lightpaths… (read more :¬†here)

Is it useful?

The deployment of translucent optical networks¬†is considered the most promising short term solution to¬†decrease costs and energy consumption in optical backbone¬†networks… (from :¬†paper)

What is its contribution?

…In¬†the T-OBS network the problem of routing and regenerator¬†placement and dimensioning (RRPD) emerges.

…RRPD is¬†a complex problem and, in order to approach it, we propose¬†to decompose it into the routing and RPD subproblems. As a¬†consequence, we provide a mixed integer linear programming¬†formulation of the routing problem and several heuristic¬†strategies for the RPD problem.

…we show¬†that the proposed RPD heuristics ensure a proper quality of¬†transmission performance whilst at the same time providing a¬†cost-effective T-OBS network architecture.¬†(from :¬†paper)

The paper presents different heuristics and respective evaluations. For the heuristics we’ve decided to implement the Biased random-key genetic algorithm (BRKGA) since its the most efficient and the one that generates the best results (generally optimal). I’ve found a very informative presentation on slideshare:

Along with the multiple animations online of the ant colony that we’ve been passionately watching today (:D), we’ve been studying well this slides and have been extending the jgap package in order to prove biased RKGA. The main differences are: the way the crossover between¬†chromosomes¬†is done for each population (with the “elite”); and the mutation process.
In the end we will also implement the ILP solution for the RPD problem (for reasonable network sizes) and test the results and the performance against our heuristic model (ILP provides optimal solutions).
Keep an eye for a future report on it ūüôā

Leave a Reply