Grid'5000 experiment

Jump to: navigation, search

Network Optimization (Networking)

Conducted by


The goal of these experiments is to test some network optimization algorithms for traffic engineering.
For these experiments we have two main requirements: -a network composed of a small number of nodes acting as an ISP "CORE" network -an overlay built on the top of this "CORE" network running network optimization algorithms
We build the "core" network by connecting some nodes through IP-tunnels and we use the QUAGGA tool in order to perform IP-routing between them.
As concern the overlay, we developed a java prototype able to run a network optimization algorithm we designed called MIRTO.
In our experiments we compare the performance of our algorithm with other two recent proposals, TEXCP and TRUMP, under different network topologies and traffic scenarios.




    Tools used



    Results allow us to conclude that dynamic yet stable traffic engineering is not only feasible but expected with rising interest by network operators.

    TEXCP is the first protocol proposed in the literature and seems to be agile and flexible enough for an implementation, while TRUMP has some limitations due to parameter tuning.
    MIRTO seems to perform better in terms of the measured performance parameters. However it has slightly slower convergence properties as it is not equation based as the other two. MIRTO moreover is simpler to implement in hardware as it does not rely on complex ECN but only on one bit congestion notification.

    Shared by:
    Last update: 2009-11-12 11:36:24
    Experiment #492