MultiPath and LoopFree Routing
How to adapt the routing plane to improve the network reliability (fast re-routing) and enhance its capacity (load balancing)?
This is the topic of my Ph.D. Thesis. We developed several contributions in the past few years. In particular, we focus on wired intra-domain link-state networks.
Our global approach relies on three main tasks:
- low complexity multipath computation algorithms (see COMNET 2010 and GIS 2009);
- preserve the loopfreeness property using a multipath hop-by-hop forwarding architecture (see ICCCN 2008);
- load balancing and fast re-routing schemes (see ICCCN 2008 and SPECTS 2008).
How to ensure a consistent routing convergence for dynamic networks in order to avoid forwarding loops?
This is the topic of one of our Ph.D. Students: François Clad. He started his thesis in 2011. This is a joint work with Jean-Jacques Pansiot (UDS), Stefano Vissicchio + Olivier Bonaventure (UCL) and Pierre François (IMDEA/CISCO systems). A brief and preliminary description of his work is also available here. We start by developing efficient algorithms to deal with any kind of modifications on a link (see TON 2013). We are currently working on adaptations of those algorithms to deal with operations on an entire node. Some illustrative slides are available here. We first develop a theoritical framework that has been accepted at ICNP 2013. We extend it in TON 2014 to deal with intermediate effects that may hamper our solution. We are now interested in practical deployment and micro-loops measurements.
Internet Measurements and Mapping
How to map, understand and analyze the properties of the Internet?
This work is mainly based on a multicast probing tool called mrinfo. We proposed several probing (e.g, mrinfo-rec - see IMC 2009 & Merlin) and mapping tools (e.g., R2AS=router-to-AS mapping - see PAM 2010 & a node degree distribution model considering L2 devices - see IMC 2010). We contribute to the topology discovery community by uploading several large datasets useful for routing evaluation purpose. Some datasets and tools are available here: Mrinfo TOOLS
We re-write mrinfo from scratch to extend its capabilities to large scale probing campaigns. We also work on a new hybrid tool, MERLIN (MEasure the Router Level of the INTernet, see NGI 2011), mixing IGMP, ICMP and ally probes. We develop a MERLIN platform (see JSAC 2011) based on a client/server architecture to overcome IGMP filtering (see GLOBECOMM 2012).
Based on such ground data, we start to investigate bipartite Internet graph models as a possible way to artificially generate IP networks (see COMNET 2013). Note that this work can be useful for routing simulation purposes.
We also investigate fingerprinting techniques and MPLS signatures (look at CCR 2012 and IMC 2013) using basic probing. Currently, and in particular in IMC 2015, we study the actual use (i.e. LDP vs. RSVPP-TE) of MPLS in real IP networks. More recently, we investigate the opportunity to reveal invisible MPLS tunnels IMC 2017. This webpage gathers our main achievements with MPLS discovery.
How to efficiently monitor the Internet at several scales?
Energy Efficient Wireless Networks
How to save energy in wireless networks?
The goal is to switch off some part of the newtork to extend its battery capacity. This is related to green networking approaches. This topic focus on distributed heuristics to approximate an MLST (Maximum Leafs Spanning Tree) in wireless networks. We have already perform a path diversity analysis in wireless sensor networks to evaluate the quantity of sensors we can turn out in a sensing only mode (to preserve the coverage, see PIMRC 2009)). We currently work on efficient diffusion schemes to achieve a good tradeoff between energy saving and routing performance (take a look to the french Master Thesis report of François Clad: and its Master Thesis slides ). In ICC 2012, we propose a performant distributed routing scheme allowing to construct an energy efficient convergecast adaptative backbone.