Multi-Homed, Multi-Hop, Wireless Transport
The predominant transport protocol on the Internet – TCP – was conceived in 1981 and is used, essentially, unchanged today. However the Internet in 30+ years later is very different: packet losses are not exclusively due to congestion in routers, for example, but also due to collisions or interference on a wireless channel. Likewise, much data traffic today is unidirectional – whereas one design tenant of TCP was bi-directionality of traffic flow. And finally, TCP was built with the assumption that devices had one, static, connection to the Internet.
Based on the observation that TCP is ill adapted today, Epizeuxis is working on characterising and quantifying the behaviours of (different flavours of) TCP when exposed to various wireless environments. Given that TCP is hard to change, Epizeuxis is also working on either developing and quantifying incrementally deployable improvements to TCP — or, to develop alternative and more adapted transport mechanisms for this environment. The objective of this research internship is to contribute to these ongoing activities, which include:
- Contribute to the development and extension of our TCP measurement tools, to extend the set of metrics that they can quantify, and to contribute to the establishment and maintenance of a dedicated real-world test platform.
- Survey, evaluate, and compare, the main TCP flavours, as well as the less-mainstream proposed extensions, alternatives and improvements to TCP — by way of real-world experiments (and, possibly, simulations where applicable).
- Design analytical models to evaluate and compare performance of transport protocols/ mechanisms in relevant scenarios.
- Contribute to development of incrementally deployable TCP adaptations, initially specifically targeting multi-hop wireless environments.
Candidate Qualifications and Skills
- The working language is English, thus a high level of English (written and oral) is required.
- A strong background in networking and communications, is required.
- The ability to work efficiently and autonomously, in a multi-cultural team, is required.
- A Strong background in mathematics and knowledge of machine learning are appreciated.
- Being at ease with programming in general, programming in C/C++ and MATLAB are appreciated.
Period & Practicalities
- Starting date: Either early Spring, or early Fall, 2017. Duration: 5-6 months.
For Further Information or To Apply: