Job opportunities

PhD positions - 2016

Multi-path in networks: detection and evaluation of performance

Advisors: Cristel Pelsser, Stéphane Cateloin, Pascal Mérindol

The first objective will be to develop measurement techniques to detect, quantify and characterize multipath in today’s Internet. To measure the performance observed by applications in a multipath environment and understand the different forms of multi-path is a second objective for this work. For this purpose, the student will develop correlation techniques in order to pinpoint the paths used by a flow and the performance of its paths.

We will subsequently draw conclusions from our measurements. Based on our observations, we will determine whether there are aspects of multipath routing that need to be improved. We may develop techniques to avoid path segments incurring degraded performance. This may require changes in the signaling of routing information, in the computation and use of the paths in order to spread traffic dynamically upon changes in load. The objective is to nevertheless provide a stable routing solution that avoids the negative effects of route oscillations.

The PhD student will elaborate methods to correlate changes in the control plane with performance degradation. If she/he observes that some actions of the network operators are detrimental to the performance, the student will propose efficient network maintenance techniques that are exempt of problems in a multi-path context.

Finally, we will study which properties can make multipath routing protocols more robust to changes than single path environments. Which physical topologies meet robustness requirements? How to design iBGP topologies?

Full subject

To apply, please send your full CV, academic record, recommendation(s) and supporting letter before may 20 2016 to pelsser@unistra.fr.

Virtualizing Heterogeneous Wireless Networks with SDN for the 5G

Direction : Fabrice Théoleyre, Antoine Gallais, Kristof Van Laerhoven

The use of Information and Communication Technologies (ICT) is becoming increasingly important. The European Telecommunication Area has been reported to bypass the aerospace and pharmaceutical domains.Telcos offer a digital connection to an increasing number of diverse devices. While the historical mobile networks were primarily designed for voice and later extended for data communication (smartphones), we now face the rise of intensive Machine-To-Machine communications. The LTE technology has recently provided a way to increase the throughput (≈300Mbps), but it is mainly dedicated for smartphones applications. In particular, such technology has not been designed to support a very large density of devices transmitting only a few data packets per day / week. The 5G networks provide a cornerstone architecture for enabling the emerging smart cities and smart buildings applications that exhibit those characteristics.

Currently, we face to concurrent technologies for IoT (LoRA, IEEE 802.11ah, IEEE 802.15.4g, IEEE 8021.5.4e-TSCH) which all exploit the same unlicensed band. Each of this technology has its own advantages and scenario of predilection : they will probably have to cohabit in the future. One of the major future challenge will consist in exploiting in symbiosis all of them. The PhD student will explore how we may use all these technologies to provide a connectivity in the Internet of Everything. First of all, he will conduct experiments on our FIT IoT-Lab (http://iot-lab.info/) platform to more accurately model their interactions, and to extract pathological situations. The second step will consist in proposing a new architecture, and new mechanisms to make them work together. We must be able to provide mechanisms to guarantee end-to-end performance : whatever the technology used for each hop, a minimum reliability or a maximum delay should be secured. Software Defined Networking (SDN) may provide interesting mechanisms to guarantee flow isolation.

PhD offer description

To apply, please send your full CV, academic record, recommendation(s) and supporting letter before may 15th 2016 to theoleyre@unistra.fr.

CIFFRE: Smart Mobility & Multi modality in the Future Complex Urban Environments

Direction : Fabrice Théoleyre, Antoine Gallais, Cristel Pelsser (ICube) & Stéphane Klein (T&S)

Our society is facing an unprecedented massive urbanization. At the time of writing, 54% of the world’s population lives in urban areas (with an 82% peak in North America), opposed to only 30% in 1950. The obvious (but not nearly trivial!) emerging challenge is how to successfully accomplish a sustainable urbanization at such level and scale. Information and Communication Technologies (ICT) are expected to play a crucial role in the sustainable de- velopment of new urban environments. The term Smart Cities has been coined to identify technology-intensive cities which provide the ability to gather, analyze, and distribute information so as to transform services offered to the citizens, improve operational efficiency, and entail better decisions at the municipal level.

Smart mobility aims at making the trips more flowing. In particular, smart cities have now to integrate new actors and new usages:

• a new age will start, with fully autonomous cars accessible on-demand;
• new entrants are catalyzing the transformation (e.g. Google, uber)
• the needs of developing flexible transportation services
• massive collaborating software solutions to reduce both the CAPEX and OPEX (e.g. blablacar, ubercommute)

A large body of techniques have been developed to compute shortest trips in road or train networks. Similarly, multi-modal transportation planning has been extensively studied in the last decades. However, the development of urban smart areas comes with unprecedented challenges:

1. web 2.0 and now 3.0 is emerging, with intensive digital interactions, virtualizing the physical world. The connection of a so large human communities empower systematic interactions and altruism. How could we exploit these interactions to make the solutions more flexible?

1. smart cities now integrate a very large collection of smart devices to monitor the traffic, regulate the traffic lights, monitoring the car park occupancy, etc. More and more local authorities develop now opendata strategies, where measures are pushed to the web to be exploited by other actors. How can this huge volume of data be used to design smarter algorithms? To regulate the traffic, optimizing both the individual (trip time), and global (e.g. pollution) interests?

1. personalization is now a requirement for most modern systems. Obviously, the proposed trips should be different for a family with a cart and a sportsman or student. Several criteria have to be considered, and the classification must be automatic: a user cannot realistically enter all his preferences for each of his requests.

PhD offer description

To apply, please send your full CV, academic record, recommendation(s) and supporting letter before may 15th 2016 to reseaux-smartmob@icube.unistra.fr.

PhD positions - 2014

Advisors : Fabrice Theoleyre, Stéphane Cateloin, Pascal Mérindol

Full subject

To apply, please send your full CV, academic record, recommendation(s) and supporting letter before may 15 2013 to cateloin@unistra.fr,merindol@unistra.fr.

PhD positions - 2013

Multipath routing and collaborative load balancing

Advisors : Thomas Noël, Jean-Jacques Pansiot, Stéphane Cateloin

Load balancing is a decision process that may be based on indicators which are local or received from other routers. The nature of these indicators should be adapted according to the context in which they are employed (wireless networks, sensor networks, code networks...). Variables relating to the quality of service could also come into consideration, for example by assigning real-time flows to one or more paths respecting specific constraints.

The main objective of this thesis is to focus on the collaboration between routers, which relies on an exchange of messages. These messages can be upstream requests such as throughput reduction ("backpressure messages"), messages allowing the admission of new flows , or disseminating information relating to the current conditions of routers (e.g. radio links quality, position and speed of a mobile node, or its battery level).

The qualitative or quantitative nature of the data exchanged, the triggers of these messages, their rate, and the action to be taken upon receipt are all behaviors to define and study. The candidate will have to evaluate the solutions he proposes, using simulation tools. Besides the quality of the distribution, convergence is a crucial point. The performance analysis will determine whether the proposed solutions are constantly evolving in an unstable state or whether tend towards an acceptable steady solution. In this case, the convergence time is an important indicator.

Full subject

To apply, please send your full CV, academic record, recommendation(s) and supporting letter before may 15 2013 to cateloin@unistra.fr.

Internet Service Provider Networks & Multi-Path Routing

Advisors : Thomas Noël, Jean-Jacques Pansiot, Pascal Mérindol

• A Thesis subject "Internet Service Provider Networks & Multi-Path Routing" is available here: Models, IP Graphs & Routing . Please contact us to apply before may 15 2013.

Networks operators must ensure a certain degree of connectivity to their customers. This induces specific patterns that structure the underlying graph, in particular the assurance of minimum redundancy. Another problem in networks of operators is their resources optimization. This objective can be implemented with multipath routing. It is one of the possible solutions to distribute the load in the network and limit congestions. The objective of this thesis is to investigate the relationships linking the graph structure of IP networks and the opportunity to efficiently develop and deploy a multipath routing protocol. The issue of valuation is then prevalent: is it easier to evaluate the valuation modification or overcome this logical overlay and then only rely on the physical structure? Is it possible to say a priori, and knowing the topological characteristics of a network, what algorithm multipath routing protocol is most effecient? Conversely, can we, given a routing algorithm, infer appropriate graph models , which favor it efficiency? Are there IP networking patterns that hamper the establishment of multiple routes or, conversely, are there any structural and logical patterns that tend to favor it?

Past positions

• 2012 - Our group proposes a PhD student position (Profile). Please, contact us for any further information.