(Autumn 2012)
Background:
A clean-slate information-centric network (ICN) architecture for Internet is being developed in the PURSUIT EU project. The system under work includes a global rendezvous service based on hierarchical DHT consisting of thousands of nodes dispersed cover different autonomous systems. The evaluation of system of this size is difficult, as router-level simulations do not scale to the whole Internet and global prototype deployment is not feasible. Therefore, we are developing our own distributed discrete event simulator tool, that can be run in a computer cluster for large-scale models. In addition, we have been collecting and combining existing data sets in order to produce a high-level, global models for Internet traffic and topology.
Description of the problem:
The work would consist of further developing the AS-level topology model, that is used in our simulation. Mainly, we are interested in augmenting the network map with high-level geographic information, that could be used to improve the estimated latencies and allow the traffic model to be made aware of the locality of the content.
Currently, we are using CAIDA’s inferred data set of AS-level relationships and an another data set can be used to assign countries to some ASes. Basically, the thesis would design heuristics to produce a complete network map based on the available partial data and the implementation of the developed model on top of our simulation environment. Secondly, our current traffic model needs to be updated to take into account the added detail in the network map by modelling the locality of traffic. Thirdly, the developed models need to validated against some measured real-world data.
A lot of the background work has already been done in another master’s thesis in the project and the main emphasis here would be on the efficient implementation. In addition, we will be writing scientific papers based on our findings.
Minimum requirements:
- Further development of heuristics to combine existing Internet topology data sets into a single comprehensive model, that incorporates geographic information for each AS.
- The modification of our existing traffic model to take into account the locality of traffic based on the geographic information provided by the network model.
- Implementation of the developed traffic and network models on top of the distributed discrete event simulator. The implementation needs to support the optimistic synchronization framework provided by the simulator tool.
- Design and implementation of the validation of the model based on measured data.
Possible extensions and alternatives:
In addition to the augmentation of the network model with geographic information, we are interested in the addition of POP-level detail to the AS-relationships map.
Work environment:
The work is conducted as a part of an international research team distributed in multiple sites. We are cooperating with, among others, the University of Cambridge and RWTH Aachen.
Skills needed:
- good spoken/written English
- programming (Python, Java, C),
- algorithm design/implementation, basic calculus and discrete math
- ability to work in a group and motivation to learn new things
Contact:
If you are interested, please contact Arto Karila, arto.karila@hiit.fi, mobile +358 500 527 452.
