Catálogo de publicaciones - libros

This paper describes a mechanism to adapt an existing wireline scheduling algorithm for a WLAN Access Point, by taking into account the error ratio affecting each flow. This enhancement is based on the idea of weighting flows according to their error ratio. Users connected over error-prone channels get their bandwidth share increased, up to a point where the overall efficiency breaks down, and the mechanism is reverted. The cost of this mechanism in terms of fairness is also addressed.

Ad-hoc wireless networks are constituted by routers enabled to establish wireless links among them in a mesh topology. This kind of networks are becoming very popular because of their potential, low cost and ease of deployment, specially when IEEE 802.11b wireless technology is used. There has been a lot of research in the design of protocols and algorithms for improving the performance of this kind of networks. However, real implementations and testbeds are scarce, and even those few well known interesting experiences are not QoS-aware and their implementation details are not consistent. This paper presents the design of a 802.11b multi-hop QoS-aware mesh network specifically designed for isolated rural environments. This proposed network is used as a testbed for driving some initial experiments that aim to measure its performance when supporting both elastic and real-time traffic. The evaluation includes guidelines for a QoS-aware deployment of the network and recommendations for further work that can improve these results.

A sensor network consists of a large number of small, low-cost devices with sensing, processing and transmitting capabilities. The sensor nodes have limited battery power; therefore energy efficiency is a critical design issue. In this paper we propose to move the sink node, called Base Station (BS) so as to decrease the energy consumption of the whole network. We present two possible strategies to move the BS: the first one minimizes the average consumed energy, while the other one minimizes the maximum transmission energy for every active sensor. To evaluate the performance of the two strategies, we compare these with the case, when the BS is deployed in a fixed position. Simulation results show that the proposed processes can reduce energy consumption, thereby significantly extending the lifetime of the entire sensor network.

The application of directional antennas in the mobile devices of wireless ad hoc networks has the potential to increase the network connectivity and capacity. However, new solutions for medium access control (MAC) are needed. This pa-per provides a survey and overview of recently proposed MAC protocols for this scenario. We summarize problems specific to this setup and categorize proposed protocols.

The 3GPP IMS subsystem, adopted also by the ITU-T/ETSI NGN initiative, proves to be a powerful framework for the provision of conferencing services. Collaborative Services, although based on multipoint conferencing, present their own set of requirements and distinguishing features. This paper tries to present an overview of the conferencing framework available in IMS-based networks, in order to analyse next the requirements and necessary extensions for advanced collaborative service provisioning in this environment. Finally, an architecture proposal for an IMS-compatible collaborative server is presented, and the undertaken validation experiments and prototypes tackled.

Structured Peer-to-Peer (P2P) networks are promising approaches for the use of P2P technologies in resource constrained environments such as mobile communication scenarios. The signaling overhead can be reduced significantly by using a hash function to map the nodes as well as the shared objects onto the same identifier space. Queries need not be flooded through the network but can be routed directly to the responsible node. This node stores references to all peers in the network providing the requested object. The mapping algorithm of the used protocol and the actual content distribution determine the number of object references one node is responsible for. This responsibility affects the maintenance traffic in structured P2P networks, as references (responsibilities) have to be shifted among nodes when a node joins or leaves the overlay. Especially in mobile scenarios a high amount of maintenance traffic heavily influences the system performance. Therefore the content distribution has great impact on the applicability of structured P2P approaches. In this paper, we present the results of our experimental analysis of content distribution in an existing large P2P network and its consequences for the structured P2P system Chord. Based on these results we evaluate the applicability of structured P2P networks in mobile scenarios.

P2P networking is a distributed model where entities play both the client and server role. One major problem addressed in this model is the discovery, searching and routing in a dynamic distributed environment. Among the different envisaged solutions, Distributed Hash Tables (DHT) are very promising. They allow the build of robust content addressable networks. Despite good theoretical performance properties, infrastructures which implement the model need a performance management framework able to monitor them in case of a concrete deployment. In this article we propose a generic performance management information model for DHTs. Our contribution uses a standard management approach based on the Common Information Model (CIM) Metric model.

Third generation Peer-to-Peer (P2P) networks are characterized by the fact, that they are based on Distributed Hash Tables (DHT). Thus, it is intended to reduce the high signaling overhead observed in unstructured P2P networks like in Gnutella. However especially if the churn rate in the considered approach is high, i.e. nodes leaving and joining the network frequently (small session duration) the signaling traffic needed to maintain the DHT structure increases considerably. In the Chord protocol, which is investigated in this work, so called fingers are used to establish shortcuts though the DHT structure to achieve a scalable routing performance. To keep these fingers up to date in regular Chord, a significant amount of signaling traffic is necessary, especially in high churn scenarios. Therefore we propose a new scheme which provides additional finger update methods without causing additional traffic.

This paper puts forward a purely statistical analysis of name server traffic captured at four different locations: two links to residential networks, and two to the Dutch academic and research institute. Analysis of the system can give insight in the use and performance of the protocol, which is helpful for future improvement. Multiple analyses can show the development of the performance over time and help create quality models. The analysis shows that a little more than 12% of all queries are not answered upon. Three quarters of the lookups are successful: they give the client the correct IP address mapping for the requested hostname. 90% is answered within 275 ms, with an average of 152 ms. In 9% of all cases, clients ask for a hostname which does not exist. At one of the locations, a client is discovered which sends queries to two DNS servers at a remarkable rate: one each 11 to 22 ms.

This paper proposes an API for Multihoming in IPv6. This API is based on the and approaches, which are being developed by the IETF multi6 Working Group. The support of Multihoming implies several actions such as failure detection procedures, reach-ability tests, re-homing procedures and exchange of locators. Applications can benefit from transparent access to Multihoming services only if per host Multihoming parameters are defined. However, more benefits could be obtained by applications if they will be able to configure these parameters. The proposed Multihoming API provides different functions to applications which can modify some parameters and invoke some functions related with the Multihoming Layer.