Catálogo de publicaciones - libros

Self-organizing systems (SOS) hold the promise of addressing many challenges in large scale distributed systems, especially in reducing the need for human intervention for configuration, recovery from failures, and performance optimization. While there are many principles for creating SOS such as minimizing dependencies between components and avoiding the use of globally shared state, we lack a systematic methodology. This talk explores how techniques from control theory and game theory might be used in combination to engineer SOS.

The Postmodern Internetwork Architecture project, a collaboration among researchers at the Universities of Kentucky, Kansas and Maryland, aims to produce a clean-slate design for a thin (inter)network layer. Design goals for this Internet layer include complete separation of routing from forwarding, avoidance of hierarchical or centrally-assigned identifiers in the forwarding plane, and provision of explicit mechanisms to support policies of the various stakeholders (specifically users and service providers).

A fundamental problem of future networks is to get fully self-organized routing protocols with good scalability properties that produce good paths in a wide range of network densities. Current approaches, geographic routing and table based routing, fail to provide very good scalability with good paths in sparse networks. We propose a method based on the discovery of connectivity between geographic regions that are self-organized in a multilevel hierarchy. The Mercator protocol builds lightweight connectivity maps in a fully decentralized manner and shows a scalable and resilient behaviour. Each node builds and maintains its own hierarchical map that summarizes connectivity information of all the network around itself using geographic regions. Link state routing is used over the multilevel connectivity graph of the map to obtain global paths. The analysis and simulation of our approach show that routing state and communication overhead grows logarithmically with network size while producing good paths.

A routing protocol is designed considering a particular environment all the time which is not possible in the case of practical ad hoc networks. Because of the uncertainty in topological rate of change, mobility model, and terrain condition, the performance is severely degraded. So the concept of assigning single routing protocol does not address the problem of most modern day mobile networks. Instead, the feedback-based routing protocol which is highly adaptable in changing environment is more suitable and this concept has been proposed in this paper. The mathematical modeling of feedback parameters have been designed and analyzed for the highly unstable networks. Some of the parameters we measure here are network connection ratio, end-to-end connectivity, packet delivery ratio, and number of nodes. Those are functionally related with the routing parameter.

In Wireless Sensor Network (WSN) applications, sensor nodes are often deployed in harsh environments. Routine maintenance, fault detection and correction is difficult, infrequent and expensive. Furthermore, for long-term deployments in excess of a year, a node’s limited power supply tightly constrains the amount of processing power and long-range communication available.

In order to support the long-term autonomous behaviour of a WSN system, a self-diagnostic algorithm implemented on the sensor nodes is needed for sensor fault detection. This algorithm has to be robust, so that sensors are not misdiagnosed as faulty to ensure that data loss is kept to a minimum, and it has to be light-weight, so that it can run continuously on a low power microprocessor for the full deployment period. Additionally, it has to be self-adapative so that any long-term degradation of sensors is monitored and the self-diagnostic algorithm can continuously revise its own rules to accomodate for this degradation. This paper describes the development, testing and implementation of a heuristically determined, robust, self-diagnostic algorithm that achieves these goals.

We discuss what kind of overlay topology should be pro-actively built before an overlay routing protocol enters a route search process on top of it.

The basic overlay structures we study are the overlay topologies, connecting every overlay node to its nearest peers.

We introduce a family of optimizations, based on a pruning rule. As flooding is a key component of many route discovery mechanisms in MANETs, our performance study focusses on the delivery percentage, bandwidth consumption and time duration of flooding on the overlay. We also consider the overlay path stretch and the overlay nodes degree as respective indicators for the data transfer transmission time and overlay resilience.

We finally recommend to optimize the K-Nearest Neighbours overlay topologies with the most selective pruning rule and, if necessary, to set a minimal bound on the overlay node degree for improving resilience.

In many scenarios, self-organization is the driving force for the use of a peer-to-peer (p2p) network. However, most current p2p networks are not truly self-organizing, as little attention has been paid on how new nodes join a p2p network, the so-called . Current p2p network protocols rely on prior-knowledge of nodes like a list of IP addresses of bootstrap servers or like a list of known peers of a p2p network. However, this kind of prior knowledge conflicts with the self-organization principle and the distributed character of p2p networks. In this paper, we present the design of a which can be used to bootstrap p2p networks of arbitrary size, even very small, private p2p networks. This bootstrap service works in today’s Internet and it can be easily integrated into existing p2p applications. We present an evaluation of the proposed bootstrapping service showing the efficiency of our approach.

In this paper we describe a self-organising and self-managing system for a Cooperative Service Provisioning (CSP) of media transport and processing services. The term cooperative is used since we assume that CSP providers as well as users offer resources to be utilised for media delivery and processing based on an Overlay Network principle. The core building block of the proposed system is a Distributed Hash Table extended with a CSP specific indexing principle and recursive search algorithm. The task of QoS constraint verification for a requested service is distributed between participating nodes. In this paper we describe CSP based on a Content Addressable Network (CAN) [1] DHT. The resulting system is evaluated based on a theoretical analysis as well as simulations.

Unstructured overlay networks are driven by simple protocols that are easy to analyze and implement. The lack of structure, however, leads to weak message delivery guarantees and poor scaling. Structured overlays impose a global overlay topology that is then maintained by all peers in a complex protocol. In contrast to unstructured approaches the structured overlays are efficient and scalable, but leave little flexibility in how their topology can be adapted to the needs of the application.

We propose a generic overlay maintenance and routing algorithm that combines the simplicity of the unstructured overlays and the scalability of the structured approaches, while allowing the application to define its own peer identifier space. The overlay topology is not explicitly defined but emerges in a self-organized way as the result of simple maintenance rules. Independently of the identifier space used, our algorithm exhibits logarithmic scaling of the average routing path length and the average node degree.

The proposed maintenance and routing algorithm is simple and places few constraints on how peers can open their connections. This together with the ability to adjust both the identifier space and the tradeoff between the path length and the node degree makes the overlay customizable in ways that are not possible in the existing approaches.

The emerging Internet and non-Internet environments have renewed interest in flexible and adaptive communication subsystems residing in end and intermediate systems, which utilise cross layer and wider network context information. To date most cross layer solutions have been very application and/or network specific, and lack re-usability. Here we propose a common architecture to support autonomic composition of functions using generic views of information derived from lower level primitives. At its heart is a distributed Information Sensing and Sharing framework. A combination of key features of this framework are the decoupling of information collection from information use, its capability to multiplex information sources, its operational independence from any specific protocol configuration, and its use outside a node context.