Sérgio Crisóstomo

Improving AODV with preemptive local route repair

This paper proposes an extension to the AODV protocol, denoted preemptive local route repair (PLRR), that aims to avoid route failures by preemptively local repairing routes when a link break is about to occur. This protocol extension resorts to AODV layer 3 connectivity information with new mobility extensions. Our proposal is to enhance nodes information concerning link stability to its neighbours resorting to HELLO messages. These messages are appended with a mobility extension containing the nodes position, motion vector and an associated timestamp. This mobility information will be used to predict the instant a link between two neighbours will break. Our proposal does not need to take into account the sender and destination location information, as other location-aided routing protocols. In this proposal, location and mobility information needs to be propagated only between neighbours. This proposal aims at improving the AODV quality of service capabilities by minimizing route failures.

Flooding the Network: Multipoint Relays versus Network Coding

Flooding is an elementary tool for information dissemination in a wide range of network scenarios, such as link state advertisements in wireless multi-hop networks and query propagation in peer-to-peer networks. Using random graph models, we compare two competing flooding techniques: multipoint relays and network coding. Our analytical results show that in the case of network coding, the number of transmissions per source message is asymptotically independent of the number of nodes. Simulation results yield further insights on the impact of topology on the performance of each flooding technique, more specifically on the required number of transmissions and the resulting end-to-end delay.

Probabilistic flooding in stochastic networks: Analysis of global information outreach

This article investigates probabilistic information dissemination in stochastic networks. The following problem is studied: A source node intends to deliver a message to all other network nodes using probabilistic flooding, i.e., each node forwards a received message to all its neighbors with a common network-wide forwarding probability @w. Question is: what is the minimum @w-value each node should use, such that the flooded message is obtained by all nodes with high probability? We first present a generic approach to derive the global outreach probability in arbitrary networks and then focus on Erdos Renyi graphs (ERGs) and random geometric graphs (RGGs). For ERGs we derive an exact expression. For RGGs we derive an asymptotic expression that represents an approximation for networks with high node density. Both reliable and unreliable links are studied.

A QoS architecture integrating mobile Ad-Hoc and infrastructure networks

Ahstract-This paper proposes an Ad-hoc QoS architecture, using as basis some concepts from SWAN, extending it to fulfill our requirements. It also proposes the modules required in the network elements and its interaction to provide optimized QoS in the ad-hoc network and integration between both networks. We also present the overall integration QoS signaling protocol and the differentiation mechanisms to address end-to-end QoS for real-time multimedia applications. Furthermore, the proposed solution addresses the use of multipath routing in order to provide load balancing and increased network reliability

Efficient Reservation-Based QoS Architecture

This paper describes a new architecture that provides end-to-end QoS support, and analyses its performance in terms of QoS guarantees and scalability of the solution. This architecture introduces a scalable per-flow signalling model, using several techniques and algorithms developed in order to minimise the computational complexity. A label switching mechanism and an efficient timer implementation were developed with the goal of reducing the signalling processing overhead at each router. The underlying architecture is based on DiffServ and the resource reservation is performed for aggregates of flows at both core and access networks. The performance results presented in this paper show that this architecture is able to support both IntServ service models in high speed networks, minimizing the processing load in each network element.

Network coding with shortcuts

Information dissemination in communication networks is a key function whose effectiveness depends both on the chosen dissemination algorithm and on the underlying network topology. A series of contributions in statistical physics has uncovered the specific topological properties of small-world networks that potentiate the spread of information, most strikingly small network diameters and large clustering coefficients. Simultaneously, the spectra of dissemination algorithms, typically based on message replication, was recently enlarged by the advent of the network coding paradigm, in which intermediate nodes are allowed to mix information flows through non-trivial linear operations. Intrigued by the dissemination of information in broadcast environments with small-world topologies, we compare the behavior of competing replication based algorithms and their network coding counterparts. We show, both analytically and through simulation, that network coding requires a smaller number of transmissions and shorter propagation delays, conjugated with impressive steadiness under distinct topological configurations.

Performance Evaluation of the RSVP Reservation Aggregation Model

This paper contains an evaluation of the RSVP Reservation Aggregation architecture, proposed by the IETF as a scalable alternative to the standard RSVP/IntServ for usage in high-speed core networks. We point out its main strengths, weaknesses and limitations, and describe our implementation of the architecture in the ns-2 simulator, including the definition of policies which are considered out of the scope of RFC3175, of which the most important is the aggregate bandwidth management policy.

An analytical model for perpetual network codes in packet erasure channels

Perpetual codes provide a sparse, but structured coding for fast encoding and decoding. In this work, we illustrate that perpetual codes introduce linear dependent packet transmissions in the presence of an erasure channel. We demonstrate that the number of linear dependent packet transmissions is highly dependent on a parameter called the width (\(\omega \)), which represents the number of consecutive non-zero coding coefficient present in each coded packet after a pivot element. We provide a mathematical analysis based on the width of the coding vector for the number of transmitted packets and validate it with simulation results. The simulations show that for \(\omega = 5\), generation size \(g = 256\), and low erasure probability on the link, a destination can receive up to \(70\%\) overhead in average. Moreover, increasing the width, the overhead contracts, and for \(\omega \ge 60\) it becomes negligible.

Evaluation of a scalable reservation-based QoS architecture

This paper presents a performance evaluation study of the scalable reservation-based QoS architecture. This architecture introduces a scalable per-flow signalling model, using techniques and algorithms developed to minimise the computational complexity, namely a label switching mechanism and an efficient timer implementation. The underlying architecture is based on DiffServ and resource reservation is performed for aggregates of flows at both core and access networks. The obtained results show that this architecture is able to provide QoS guarantees, irrespectively of the behavior of other flows in the same and in different classes, maximizing the network resource utilization. Based on the performance evaluation, we can state that this architecture is able to support service classes with strict and soft QoS guarantees in high-speed networks.

SRBQ and RSVPRAgg: A Comparative Study

This paper presents a comparative evaluation of the Scalable Reservation-Based QoS (SRBQ) and the RSVP Reservation Aggregation (RSVPRAgg) architectures, both designed to provide QoS levels similar to RSVP/IntServ without the scalability concerns that prevent its usage in high-speed core networks. The comparative analysis, based on simulation results, shows that SRBQ provides the same QoS guarantees of RSVPRAgg, with significantly increased network resource utilisation and a small penalty in signalling processing overhead. This stems from the fact that although based on end-to-end reservations, SRBQ makes use of techniques and algorithms that reduce the computational complexity of signalling processing, increasing its scalability.