• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 1
  • Tagged with
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

Control de congestión multipunto en redes IP y ATM. Diseño de protocolos de transporte multipunto fiable

Solera Delgado, Marta 13 November 2006 (has links)
Las comunicaciones multipunto ofrecen, tanto a usuarios como a proveedores, mayor eficiencia, permitiendo desarrollar e implantar nuevos servicios. Para poder desplegarlos adecuadamente sobre la redes de comunicaciones, es necesario contar con protocolos adecuados a todos los niveles.El soporte de conexiones multipunto, que es inmediato en muchas redes de ordenadores, por la existencia de un medio compartido, no lo es en una red ATM o en una red IP. En este tipo de redes, ofrecer una comunicación multipunto requiere de mecanismos complejos que coordinen y controlen la transmisión de información entre las fuentes y los receptores. Esta tesis doctoral estudia las comunicaciones punto a multipunto sobre las redes ATM e IP. En concreto, se diferencian dos objetivos: · Estudiar, analizar y proponer un control de flujo punto a multipunto en la categoría de servicio Available Bit Rate (ABR) en redes ATM.· Diseñar, analizar y simular un protocolo de transporte punto a multipunto fiable con control de congestión de tasa única para redes IP.El control de flujo de ABR fue diseñado para comunicaciones punto a punto. Para el caso multipunto, los conmutadores deben desarrollar mecanismos que limiten y agreguen el tráfico de realimentación. Se ha desarrollado un algoritmo de consolidación que asegura que la agregación de la información de realimentación se realiza de forma correcta, mejorando la convergencia de propuestas anteriores. Este mecanismo se ha modelado matemáticamente, y se ha validado mediante simulación.En cuanto a las comunicaciones multipunto en Internet, en este trabajo se ha desarrollado un protocolo de transporte punto a multipunto fiable, RCCMP, diseñado para ser escalable, fiable y con un control de congestión de tasa única que comparta el ancho de banda equitativamente con TCP.El control de congestión ha sido planteado como una parte esencial del protocolo, y no como ocurre en muchas propuestas, como un componente adicional que debe ser ajustado a un protocolo de transporte. En RCCMP se combinan los objetivos de regular la tasa de transmisión y de conseguir una comunicación fiable, con el fin de simplificar y limitar el número de confirmaciones negativas que se envían desde los receptores. Para la evaluación de las prestaciones de RCCMP, se ha implementado el protocolo en el simulador ns-2. Se ha modelado el caudal de este protocolo de transporte multipunto en función de la tasa de pérdidas y del tiempo de ida y vuelta. La principal contribución de nuestro modelo radica en la caracterización del caudal ante cambios de representante. También, se ha desarrollado un método de análisis para estimar el ancho de banda consumido por cualquier protocolo de transporte. La principal diferencia con otros trabajos es que éstos se centran únicamente en el coste de los procesos del control de errores.Para mejorar el rendimiento de RCCMP, se ha definido y simulado otro protocolo de transporte punto a multipunto fiable de tasa única, RVCMP, que incluye un control de congestión que emula al de la implementación TCP Vegas. / Multicast communications are profitable for service providers increasing intermediate node performance and reducing traffic in their networks. In the other side, multicast benefits also the users, who are able to enjoy collaborative applications and other multicast applications. IP and ATM network were designed to support point to point communications. The new multicast generation applications such multimedia conference, shared workspace, distance learning introduce new requirements in data transmission. There are at least two problems that differentiate between unicast and multicast control schemes in ATM networks with ABR service. First, there is the problem of feedback implosion. The volume of feedback traffic returning to the source increases proportional to the number of destinations. Second, there is the problem of consolidation noise. It can occur when feedback from some leaves is not always received in a timely fashion at the time when RM cells need to be returned by the branch point. One of the proposals to provide multicast communications is to extend unicast traffic management control methods. One of its practical realization schemes is the extension of Enhanced Proportional Rate Control Algorithm (EPRCA) for point to multipoint connections. That proposal suffers from consolidation noise, in order to solve this drawback, a new algorithm is proposed. An analytical approach is used to quantitatively evaluate their performance by using first-order fluid approximation method.About IP multicast communications, we propose a new protocol, called RCCMP (Reliable Congestion Controlled Multicast Protocol). It has been designed to be simple, scalable (NAK suppression), reliable and TCP-friendly. The congestion control is a central part of the protocol, where the feedback of the worse receiver is used to control a transmission window in a TCP-like fashion. The scalability issue is addressed with an exponential timer scheme that is also used to estimate the number of receivers involved in the communication. The protocol neither needs support from network elements nor maintains state information dependent of the number of receivers. This protocol has been developed in ns-2 (network simulator-2) for validating.We develop a simple analytic characterization of the steady-state send rate as a function of loss rate and round trip time (RTT). The main contribution is that our model captures the representative changes. Also, we provide a new approach to estimate the link usage of multicast reliable transport protocols. It can be used as a benchmark to evaluate their scalability. We have chosen, as a case study, a multicast reliable transport protocol called RCCMP. The link cost is due to data, retransmitted data, and characteristic packets of multicast protocols such as control packets: control packets for estimating the number of receivers involved in the session and for getting multicast reliability. For improving performance, we present RVCMP (Reliable Vegas Congestion controlled Multicast Protocol) that has been designed to be simple, scalable, reliable and TCP-friendly. The congestion control developed is a single-rate scheme where the feedback of the worst receiver is used to control a transmission window in a TCP Vegas-like fashion. The proposal takes the advantage from TCP Vegas of operating without inducing packet losses as a signal that there is congestion in the network to achieve a better performance. To evaluate the benefits of Vegas-like congestion control, the performance of RVCMP is compared to an analogous protocol that is based on a TCP Reno congestion control, RCCMP.

Page generated in 0.1368 seconds