An active network is a computer network in which new protocols can be installed at run-time in any node within the network. For example, the deployment of Internet multicast technology has been slow because service providers have been reluctant to upgrade and reconfigure their routing nodes. Under the active network scheme, users who desire multicast services can have the service automatically installed without any direct intervention by the user or the provider. One major question in realizing active networks is how the code for the new active library can be found, or resolved, and retrieved. A model of the resolution and retrieval mechanisms is the major focus of this research. To validate the model, a proof-of-concept experimental system that realizes a simple active network architecture was developed. An active library resolution service model, suitable for a global Internet, was investigated using this experimental platform and a simulation system. The two protocol components that were built and studied are the active transport protocol and the active library resolution protocol. The experimental and simulation systems were used to evaluate the extensibility, overhead, resolution time, scalability, and policy constraint support of the service. Extensibility and policy constraint support are an integral part of the proposed design. For libraries located on servers that are at most ten hops away from the requesting source, the resolution time is under 2.6 seconds. Simulations of networks of different sizes and with different error rates exhibit linear resolution time and overhead characteristics, which indicates potential scalability. Behavior under high loss rates showed better than expected performance. The results indicate that the library resolution concept is feasible and that the proposed strategy is a good solution. / Ph. D.

Telecommunication service providers have recently begun to offer ubiquitous access to packetised data. As a result, the Internet is not limited to computers that are physically connected but is also available to users that axe equipped with mobile devices. This ubiquitous access fuels the growth and the usage of the Internet even further, and thus the realisation of dynamic Internet. With the realisation of the dynamic Internet, increasing support is needed for Internet protocol (IP) and transmission control protocol (TCP) over wireless/mobile networks.Two areas of interest in this thesis are unicast and multicast routing in connectionless and connection-oriented networks. To address the problems of routing protocols in mobile computing environments, the active networks (ANs) paradigm is employed. ANs provide an alternative paradigm to solving network problems and comprise programmable network elements that allow enhancement of existing protocols and the execution of active protocols which run for the duration of the communication session.This thesis investigates the viability and advantages of ANs when applied to routing in mobile communications. Two new AN-based protocols, for IP and asynchronous transfer mode (ATM) networks, that address the problems of multicast routing with mobile group members are outlined. The Internet Engineering Task Force (IETF) mobile IP has been augmented with active programs in order to enhance its operation further. Also, a novel model for rerouting connections in ATM networks is presented.Results of extensive simulation studies comparing performances of conventional as well as some recently proposed protocols with those of AN-based protocols are presented. The results obtained from these simulation studies show that AN-based protocols have the following benefits: (i) efficient adaptation to mobility, (ii) reduced signaling overheads, ++ / (iii) high reuse of allocated network states, (iv) extensibility, (v) network topology independence, and (vi) scalability. The aforementioned points are crucial in mobile environments where states at routers (switches) are frequently updated due to mobility. It was shown that ANs provide the most benefits to protocols that maintain states within the network, for example connection-oriented and multicast protocols. AN-based protocols enable fast and efficient update of the states maintained at the routers/switches without incurring excessive signaling overheads. Moreover, part of a connection or multicast tree can be updated iteratively with the use of ANs, resulting only in modifications to routers (switches) that are affected by host migration. A model for deploying active programs that is coupled with the protocol operation is also demonstrated. Implementation of such a model eliminates the need for strategic positioning of active services.

No description available.

No description available.

Ενεργά λέμε τα δίκτυα τα οποία επεξεργάζονται και τα περιεχόμενα (και όχι μόνο την επικεφαλίδα) των πακέτων που μετάγουν. Επικεντρωνόμαστε στα ενεργά δίκτυα ενθυλάκωσης, όπου ο προς εκτέλεση κώδικας συμπεριλαμβάνεται στα μεταγώμενα πακέτα, σε αντιδιαστολή με τους προγραμματιζόμενους μεταγωγείς. Παρουσιάζεται αναλυτικά το Active Networks Encapsulation Protocol (ANEP). Παρουσιάζονται δικτυακές εφαρμογές στις οποίες τα ενεργά δίκτυα βελτιώνουν την απόδοση της εφαρμογής και ταυτόχρονα μειώνουν τις απαιτήσεις σε δικτυακούς πόρους. Ακολούθως εστιάζουμε στην \"Ενεργή Αξιόπιστη Πολλαπλή Μετάδοση\", ένα πρωτόκολλο αξιόπιστης πολλαπλής μετάδοσης το οποίο χρησιμοποιεί την ενεργή μεταγωγή για να διαχειριστεί την ανάκτηση απωλειών πακέτων εντός του δικτύου (καταστολή NACK, λανθάνουσα μνήμη για πακέτα διόρθωσης, πολλαπλή μετάδοση περιορισμένης εμβάλειας). Παρέχονται αποτελέσματα προσομοιώσεων που υποστηρίζουν την θέση ότι ακόμα και με μικρό ποσοστό ενεργών κόμβων, ένα ενεργό δίκτυο μπορεί να βελτιώσει ουσιαστικά τις επιδόσεις της εφαρμογής και να μειώσει ταυτόχρονα την χρήση εύρους ζώνης. Κλείνοντας, κάνουμε κάποια τελικά σχόλια και εξάγουμε συμπεράσματα σχετικά με το υψηλό κόστος εγκατάστασης και συντήρησης των ενεργών δίκτυων, και πως αυτό αντιδιαστέλλεται με τα πλεονεκτήματα των τελευταίων σε σχέση με τις επιδόσεις των εφαρμογών και την χρήση των δικτυακών πόρων. / Active Networks are networks consisting (at least partially) of active nodes. A node is active if it doesn’t only processes a packet’s header in order to route it, but is also able to evaluate and process the packet’s payload. There are two kinds of active networks, depending on whether they are based on programmable switches or on capsules which bundle code together with the data. This dissertation focuses on the latter. The operational model of an active network of this kind comprises code execution models, network node management models and resource allocation policies. The Active Networks Encapsulation Protocol (ANEP) sets the mechanism for defining the platform required to evaluate the code that is encapsulated in the packet, as well the nodes’ behaviour when they do not support the required platform (drop the packet or simply forward it). This mechanism provides active networks with the flexibility to operate even when a very small percentage of the network’s node is actually active. There are various situations and where active networks can make better use of network resources. There are, for example, applications where different users might make similar, but different, requests resulting in unnecessary bandwith consumption when supported by conventional caching mechanisms. Active networks can provide smart caches that will dynamically synthesize pages from data cached by previous requests. Another situation where active networks can improve network performance is network applications like tele-conference, that depend heavily on new network services. Active networks allow for the faster deployment of new network services that enhance network speed and security and rationalise bandwidth usage through, for example multicast. Furthermore, active networks can support specialised applications, like for example on-line auctions, with custom-made network services. It is important to note that when measuring network performance, one should focus onto the network application’s performance, rather than network per-packet metrics like throughput and latency. In other words, intranetwork processing might increase both packet size and latency, but will improve the application’s end-to-end performance and reduce total network load. The protocols for three innovative network applications are presented: active reliable multicast, auctions over the network and remote sensor merging. For each of these we present network services that can be easily implemented and deployed in active networks to improve application performance. Finally, a more detailed analysis (by means of simulation) of an active reliable multicast protocol is presented. Active networks achieve two ends: on the one hand they push the idea of a network proxy to its logical end by effectively turning all network elements into smart proxies that provide caching, filtering, NACK suppression and other services. On the other hand they carry out part of the computation inside the network, bringing it closer to the data sources. When the computation is, for example, data merging this is beneficial to both the application and the network resources. This, however, can only be achieved at a cost. First of all in hardware, since network elements have to be upgraded from simple routers to full-blown computers capable of supporting Java and scripting languages. But also in latency, since packets have to undergo much more complex processing along the way that simple routing. In the applications presented here the costs associated with active transport are counter-balanced by the advantages the latter has to offer to the application as well as to the network. The bet that active networks have to win in order to get widely accepted, is to have enough active application protocols developed that their installation and maintenance cost can be justified.

Les expectatives dels usuaris en relació a la quantitat i qualitat del serveis de xarxa estan creixent ràpidament. En canvi, desenvolupar e implantar nous serveis de xarxa (serveis que operen a nivell de xarxa) seguint el procés d'estandardització és massa lent i no pot satisfà les expectatives.Les xarxes actives i programables van ser proposades per acomodar la ràpida evolució de noves tecnologies i accelerar la implantació de serveis sofisticats. Les xarxes actives permeten que tercers (usuaris finals, operadors i proveïdors de servei) introdueixin serveis específics per aplicacions (en forma de codi) dins la xarxa. D'aquesta manera, les aplicacions poden fer servir aquests serveis per obtenir el suport necessari de la xarxa en termes de comportament per exemple.Tanmateix, les tecnologies de xarxes actives i programables introdueixen una complexitat addicional als elements de xarxa que ha de ser tractada pel sistema de gestió. Alguns exemples d'aquesta complexitat addicional són la necessitat de suportar nous serveis introduïts dinàmicament a la xarxa o la gestió de xarxes actives virtuals. Aquestes xarxes poden ser creades dins una infrastructura de xarxa programable per satisfer les creixents necessitats de control i particularització que els clients imposen a les xarxes. A més, la probable implantació progressiva de les xarxes actives i programables dins les xarxes IP actuals afegeix un altre requisit important al pla de gestió: aquest ha de ser capaç de suportar tecnologies de xarxa heterogènies (passives, actives i programables).La solució proposada en aquesta tesi millora els sistemes de gestió basats en polítiques amb conceptes de les tecnologies de xarxes actives i programables per satisfer tots els requisits abans esmentats; assolint, d'aquesta manera, una solució sòlida per la gestió de xarxes actives i programables.Finalment, l'arquitectura per la gestió de xarxes actives basada en polítiques (MANBoP) que proposem ha estat dissenyada per poder ser instanciada a qualsevol nivell de gestió (xarxa, sub-xarxa o element). A més, diferents instàncies es poden agrupar fàcilment per crear una infrastructura de gestió. Per exemple, una instància MANBoP de nivell de xarxa pot treballar sobre instàncies de nivell de element, o vàries instàncies de nivell de sub-xarxa poden ser creades per treballar en paral·lel, cadascuna gestionant una regió geogràfica de la xarxa diferent. L'objectiu d'aquest atribut de l'arquitectura és facilitar la tasca de creació d'una infrastructura de gestió. D'aquesta manera, els operadors de xarxa poden crear la infrastructura de gestió que més els convingui segons els seus objectius de negoci i reduir així els costs de gestió. / -RESUMENLas expectativas de los usuarios en relación con la cantidad y calidad de los servicios de red están creciendo rápidamente. En cambio, desarrollar e implantar nuevos servicios de red (servicios que operan al nivel de red) siguiendo los procesos de estandarización es demasiado lento y no colma las expectativas.Las redes activas y programables fueron propuestas para acomodar la rápida evolución de las nuevas tecnologías y acelerar la implantación de nuevos servicios más sofisticados. Las redes activas permiten que terceros (usuarios finales, operadores o proveedores de servicio) introduzcan servicios específicos para aplicaciones (en forma de código) dentro de la red. De esta forma, las aplicaciones pueden utilizar estos servicios para obtener el soporta que necesitan de la red en términos, por ejemplo, de comportamiento.Sin embargo, las tecnologías de redes activas y programables introducen una complejidad adicional en los elementos de red que debe ser tratada por el sistema de gestión. Algunos ejemplos de esta complejidad adicional son la necesidad de soportar nuevos servicios introducidos dinámicamente en la red o la gestión de redes activas virtuales. Éstas pueden ser creadas dentro de una infraestructura de red programable para satisfacer las necesidades crecientes de control y particularización que los clientes imponen sobre las redes. Además, la probable progresiva implantación de la redes activas y programables en la redes IP actuales añade otro importante requisito al plano de gestión: éste tiene que ser capaz de soportar tecnologías de red heterogéneas (pasivas, activas y programables).La solución propuesta en esta tesis mejora los sistemas de gestión basados en políticas con conceptos de las tecnologías de redes activas y programables para satisfacer todos los requisitos enumerados anteriormente, consiguiendo, de esta forma, una solución sólida para la gestión de redes activas y programables.Finalmente, la arquitectura de gestión de redes activas basada en políticas (MANBoP) que proponemos ha sido diseñada para poder ser instanciada en cualquier nivel de gestión (red, sub-red o elemento). Además, diferentes instancias pueden ser agrupadas fácilmente para crear una infraestructura de gestión. Por ejemplo, una instancia MANBoP de nivel de red puede trabajar sobre instancias de nivel de elemento, o varias instancias de nivel de sub-red pueden ser creadas para trabajar en paralelo sobre diferentes regiones geográficas de la red. El objetivo de esta característica de la arquitectura es facilitar la creación de una infraestructura de gestión para que los operadores de red puedan crear la que más les convenga según sus objetivos de negocio, reduciendo así los costes de gestión. / User expectations of the range and quality of network services are growing rapidly. However, developing and deploying new network services (i.e. services that operate on the network layer), through best practice and standardization, is too slow and cannot match the steps of expectations. Active and programmable networks were proposed to accommodate the rapid evolution of new technologies and accelerate the deployment of new sophisticated services. Active networks (AN) enable third parties (end users, operators, and service providers) to inject application-specific services (in the form of code) into the network. Applications are thus able to utilize these services to obtain required network support in terms of, e.g. performance; that is, applications are now becoming network-aware. Nonetheless, active and programmable networking technologies introduce additional complexity in network elements that must be handled by the management architecture. Examples of this additional complexity are the need of coping with new services dynamically introduced in the network, or the management of virtual active networks (VANs). VANs might be created over a programmable network infrastructure to satisfy increasing requirements for the control and customization of resources that customers impose on networks. Furthermore, the likely progressive deployment of active and programmable routers in today's IP networks introduces another important requirement in the management plane; that is, it has to be able to cope with heterogeneous network technologies, i.e., passive, active and programmable.The solution proposed in this thesis enhances a policy-based management system with active networking technology concepts to fulfill the above-mentioned requirements, thus achieving a sound solution for the management of active and programmable networks.In addition, the Management of Active Networks Based on Policies (MANBoP) framework proposed is designed to be instantiated at any management level (i.e. network, sub-network or element). Moreover, different instances can be easily set up jointly to create a management infrastructure, e.g., a network-level MANBoP instance can be set up over element-level instances, or several subnetwork-level instances can work in parallel each one managing a different geographical region of the network, etc. The aim of this framework property is to ease the management infrastructure creation task, thus allowing network operators to adapt the management infrastructure to their business needs, resulting in a reduction of management costs.

High-throughput whole-genome biological assays are highly intricate and difficult to interpret. The molecular interaction networks generated from evaluation of those experiments suggest that cellular functions are carried out by modules of interacting molecules. Reverse-engineering the modular structure of cellular interaction networks has the promise of significantly easing their analysis. We hypothesize that: • cellular wiring diagrams can be decomposed into overlapping modules, where each module is a set of coherently-interacting molecules and • a cell responds to a stress or a stimulus by appropriately modulating the activities of a subset of these modules. Motivated by these hypotheses, we develop models and algorithms that can reverse-engineer molecular modules from large-scale functional genomic data. We address two major problems: 1. Given a wiring diagram and genome-wide gene expression data measured after the application of a stress or in a disease state, compute the active network of molecular interactions perturbed by the stress or the disease. 2. Given the active networks for multiple stresses, stimuli, or diseases, compute a set of network legos, which are molecular modules with the property that each active network can be expressed as an appropriate combination of a subset of modules. To address the first problem, we propose an approach that computes the most-perturbed subgraph of a curated pathway of molecular interactions in a disease state. Our method is based on a novel score for pathway perturbation that incorporates both differential gene expression and the interaction structure of the pathway. We apply our method to a compendium of cancer types. We show that the significance of the most perturbed sub-pathway is frequently larger than that of the entire pathway. We identify an association that suggests that IL-2 infusion may have a similar therapeutic effect in bladder cancer as it does in melanoma. We propose two models to address the second problem. First, we formulate a Boolean model for constructing network legos from a set of active networks. We reduce the problem of computing network legos to that of constructing closed biclusters in a binary matrix. Applying this method to a compendium of 13 stresses on human cells, we automatically detect that about four to six hours after treatment with chemicals cause endoplasmic reticulum stress, fibroblasts shut down the cell cycle far more aggressively than fibroblasts or HeLa cells do in response to other treatments. Our second model represents each active network as an additive combination of network legos. We formulate the problem as one of computing network legos that can be used to recover active networks in an optimal manner. We use existing methods for non-negative matrix approximation to solve this problem. We apply our method to a human cancer dataset including 190 samples from 18 cancers. We identify a network lego that associates integrins and matrix metalloproteinases in ovarian adenoma and other cancers and a network lego including the retinoblastoma pathway associated with multiple leukemias. / Ph. D.