ISAIA, P. and GUAN, L., 2017. Distributed mininet placement algorithm for fat-tree topologies. Presented at the IEEE 25th International Conference on Network Protocols (ICNP), Toronto, Canada, 10th-13th October 2017.
Distributed Mininet implementations have been extensively used in order to overcome Mininet’s scalability issues. Even though they have achieved a high level of success, they still have problems and can face bottlenecks due to the insufficient placement techniques. This paper proposes a new placement algorithm for distributed Mininet emulations with optimisation for Fat-Tree topologies. The proposed algorithm overcomes possible bottlenecks that can appear in emulations due to uneven distribution of computing resources or physical links. In order to distribute the emulation experiment evenly, the proposed algorithm assigns weights to each available machine as well as the communication links depending on their capabilities. Also, it performs a code analysis and assigns weights to the emulated topology and then places them accordingly. Some noticeable results of the proposed algorithm are the decrease in packet losses and jitter by up to 86% and 68% respectively. Finally, it has achieved up to 87% reduction in the standard deviation between CPU usage readings of experimental workers.