Abstract

 

Wednesday, July 5th, 2017. Time: 12h.

Place: O.C. Zienkiewicz Conference Room, C1 Building, UPC Campus Nord, Barcelona.

ABSTRACT

Discrete Exterior Calculus is a recent method to discretize partial differential equations using concepts from Differential Geometry and Topology, which allow rewriting equations in an intrinsic manner.

The main steps in such an approach are as follows:

1. Write out the PDE in terms of smooth differential operators.
2. Replace the smooth domain, variables, and operators by their DEC counterparts.
3. Use DEC-based analysis to set up an appropriate linear system of equations.

The resulting linear system can then be solved by existing numerical methods to approximate solutions to the PDE.

SPEAKER CV

Dr. Rafael Herrera Guzmán obtained his Degree in Mathematics at the UNAM in 1993 with the medal “Gabino Barreda”. He is PhD in Mathematics by the Oxford University, England, in 1998. He is a researcher at CIMAT from 2005 and member of the SNI Level II within the Physics, Mathematics and Earth Sciences Area.

Previously, Dr. Herrera has worked at Yale University, University of California at Riverside, Princeton University and City University of New York. He has been awarded with the “Guggenheim Fellowship” (2002-2003) and a “Research Fellowship” by the Japan Society for the Promotion of Science (2005-2006).

Abstract

The Internet is a wide network of computers in constant evolution. Each year, more and more organizations are connected to this worldwide network. Each of them has its own structure and administration that are not publicly revealed for economical, political, and security reasons. Consequently, our perception of the Internet structure, and more specifically, its topology, is incomplete. In order to balance this lack of knowledge, the research community relies on network measurements. Most of the time, they are performed based on the well-known tool traceroute. However, in practice, an operator may privilege other technologies than IP to forward packets inside its network. MultiProtocol Label Switching (MPLS) is one them. Even if it is heavily deployed by operators, it has not been really investigated by researchers. Prior to this thesis, only two studies focused on the identification of MPLS tunnels in traceroute data. Moreover, while one of them does not take all possible scenarios into account, the other lack of precision in some of its models. In addition, MPLS tunnels may hide their content to traceroute. Topologies inferred from such data may thus contain false links or nodes with an artificially high degree, leading so to biases in standard graph metrics used to model the network. Even if some researchers already tried to tackle this issue, the revelation of hidden MPLS devices in traceroute data is still an open question. This thesis aims at characterizing MPLS in two different ways. On the one hand, at an architectural level, we will analyze in detail its deployment and use in both IPv4 and IPv6 networks in order to improve its state-of-the-art view. We will show that, in practice, more than one IPv4 trace out of two crosses at least one MPLS tunnel. We will also see that, even if this protocol can simplify the internal architecture of transit networks, it also allows some operators to perform traffic engineering in their domain. On the other hand, MPLS will be studied from a measurement point of view. We will see that routers from different manufacturers may have distinct default behaviors regarding to MPLS, and that these specific behaviors can be exploited to identify MPLS tunnels during traceroute measurements. More precisely, we will focus on new methods able to infer the presence of tunnels that are invisible in traceroute outputs, as well as on mechanisms to reveal their content. We will also show that they can be used in order to improve the inference of Internet graph properties, such as path lengths and node degrees. Finally, these techniques will be integrated into Trace the Naughty Tunnels (TNT), a traceroute extension able to identify all types of MPLS tunnels along a path towards a destination. We will prove that this tool can be used in order to get a detailed quantification of MPLS tunnels in the worldwide network. TNT is publicly available, and can therefore be part of many future studies conducted by the research community. Internet est un immense réseau informatique en constante évolution. Chaque année, de plus en plus d’organisations s’y connectent. Chacune d’elles est gérée et administrée indépendamment des autres. En pratique, l’architecture interne de leur réseau n’est pas rendue publique pour des raisons politiques, économiques, ou de sécurité. Par conséquent, notre perception de la structure d’Internet, et plus particulièrement de sa topologie, est incomplète. Afin de pallier ce manque de connaissance, la communauté de la recherche s’appuie sur des mesures de réseau. La plupart du temps, elles sont réalisées avec l’outil traceroute. Cependant, des technologies autres que IP peuvent être privilégiées pour transférer les paquets dans un réseau. MultiProtocol Label Switching (MPLS) est l’une d’entre elles. Même si cette technologie est largement déployée dans Internet, elle n’est pas bien étudiée par les chercheurs. Avant cette thèse, seulement deux travaux se sont intéressés à l’identification d’MPLS dans les données collectées avec traceroute. Alors que le premier ne prend pas en compte tous les scénarios possibles, le second propose des modèles qui manquent de précision. De plus, les tunnels MPLS peuvent dissimuler leur contenu à traceroute. Les topologies inférées sur base de ces données peuvent donc contenir de faux liens, ou des noeuds avec un degré anormalement élevé. Les différentes modélisations d’Internet qui en résultent peuvent alors être biaisées. Aujourd’hui, la question de la révélation des routeurs MPLS qui sont invisibles dans les données de mesure n’est toujours pas résolue, même si certains chercheurs ont déjà proposé quelques méthodes pour y parvenir. Cette thèse a pour but de caractériser MPLS de deux manières différentes. Dans un premier temps, au niveau architectural, nous analyserons en détail son déploiement et son utilisation dans les réseaux IPv4 et IPv6 afin d’améliorer l’état de l’art. Nous montrerons qu’en pratique, plus d’une trace IPv4 sur deux traverse au moins un tunnel MPLS. Nous découvrirons également que bien que ce protocole peut être utilisé pour simplifier l’architecture interne des réseaux de transit, il peut aussi être déployé pour la mise en place de solutions d’ingénierie de trafic. Dans un second temps, MPLS sera étudié d’un point de vue mesure. Nous verrons que les comportements par défaut liés au protocole varient d’un fabricant de routeur à l’autre, et qu’ils peuvent être exploités afin d’identifier les tunnels MPLS dans les données traceroute. Plus précisément, nous découvrirons de nouvelles méthodes capables d’inférer la présence de tunnels invisibles avec traceroute, ainsi que de nouvelles techniques pour révéler leur contenu. Nous montrerons également qu’elles peuvent être utilisées afin d’améliorer la modélisation d’Internet. Pour terminer, ces techniques seront intégrées à Trace the Naughty Tunnels (TNT), une extension de traceroute qui permet d’identifier tous les types de tunnels MPLS le long du chemin vers une destination. Nous prouverons que cet outil peut être utilisé pour obtenir des statistiques détaillées sur le déploiement d’MPLS sur Internet. TNT est disponible publiquement, et peut donc être librement exploité par la communauté de la recherche pour de multiples futures études.

Abstract

Charging stations are becoming an important part of the modern urban infrastructure. This paper describes a concept of a standalone cost-effective medium power charging station dedicated for electric bus charging. It is based on a voltage-source active rectifier operating in both current and voltage mode supplied by low voltage power grid. Appropriate model and simulation of the station at the power level of 300kW are provided.

Abstract

Many scientific applications deal with data from a multitude of different sources, e.g., measurements, imaging and simulations. Each source provides an additional perspective on the phenomenon of interest, but also comes with specific limitations, e.g. regarding accuracy, spatial and temporal availability. Effectively combining and analyzing such multimodal and partially incomplete data of limited accuracy in an integrated way is challenging. In this work, we outline an approach for an integrated analysis and visualization of the atmospheric impact of volcano eruptions. The data sets comprise observation and imaging data from satellites as well as results from numerical particle simulations. To analyze the clouds from the volcano eruption in the spatiotemporal domain we apply topological methods. We show that topology-related extremal structures of the data support clustering and comparison. We further discuss the robustness of those methods with respect to different properties of the data and different parameter setups. Finally we outline open challenges for the effective integrated visualization using topological methods. Document type: Part of book or chapter of book

Abstract

Forecasting Internet traffic is receiving an increasing attention from the computer networks domain. Indeed, by improving this task efficient traffic engineering and anomaly detection tools can be developed, leading to economic gains due to better resource management. This paper presents a Neural Network (NN) approach to predict TCP/IP traffic for all links of a backbone network, using both univariate and multivariate strategies. The former uses only past values of the forecasted link, while the latter is based on the neighbor links of the backbone topology. Several experiments were held by considering real-world data from the UK education and research network. Also, different time scales (e.g. every ten minutes and hourly) were analyzed. Overall, the proposed NN approach outperformed other forecasting methods (e.g. Holt-Winters). R&D Algoritmi centre Document type: Part of book or chapter of book

Abstract

Proceeding of: IEEE International Conference on Communications (ICC 2018) Enterprise Networking has a strong set of requirements in terms of resiliency, reliability and resources usage. With current approaches being based on monolithic and expensive infrastructures using dedicated overlay links, providers are moving to more economical hybrid solutions that encompass private dedicated links with public/regular Internet connections. However, these usually rely on complex, hardware-dependent and/or proprietary Traffic Engineering (TE) solutions, which are computationally costly, in particular for the forwarding nodes. In this paper, we propose SEMPER: a lightweight TE solution based on MP-TCP that, in contrast to other TE solutions, moves the complexity to the endpoints of the connection, and relieves the forwarding elements from complex operations or even maintaining state. As our evaluation shows, SEMPER efficiently makes use of all available paths between the endpoints while maintaining fairness, and properly adapts to variations on the available capacity. This work has been partly supported by the H2020 5GMoNArch project (grant agreement 761445), and by the Madrid Regional Government through the TIGRE5-CM program (S2013/ICE-2919).

Abstract

International audience; It is shown that the queuing delay involved in the congestion control algorithm is state-dependent and does not depend on the current time. Then, using an accurate formulation for buffers, networks with arbitrary topologies can be built. At equilibrium, our model reduces to the widely used setup by Paganini et al. Using this model, the delay-derivative is analyzed and it is proved that the delay time-derivative does not exceed one for the considered topologies. It is then shown that the considered congestion control algorithm globally stabilizes a delay-free single buffer network. Finally, using a specific linearization result for systems with state-dependent delays from Cooke and Huang, we show the local stability of the single bottleneck network.