Catálogo de publicaciones - libros

Existing wireless networks have little in common, as they are designed around vendor-specific devices that use specific radio access technologies to provide particular functionality. Next generation networks seek to integrate wide-area and local-area wireless systems in order to provide seamless services to the end user. This would provide freedom of movement between indoor/outdoor and metropolitan/enterprise coverage while maintaining continuity of applications experience. Seamless Mobility is an experiential architecture, predicated on providing mechanisms that enable a user to accomplish his or her tasks without regard to technology. This paper examines how autonomic mechanisms can satisfy some of the challenges in realizing seamless mobility solutions.

We present qualitative arguments concerning the probable infection pattern in a directed graph under the (weak or strong) influence of the outside world. This question is relevant for real computer viruses, which spread by following the (logical) directed links formed by address lists. Our arguments build on previous work in two (seemingly unrelated) areas: epidemic spreading on undirected graphs, and eigenvectors of directed graphs as applied to Web page ranking. More specifically, we borrow a recently proven result (used to design a ’sink remedy’ for Web link analysis) and use it to argue for a : that the effects of the outside world will not appear in the pattern of infection until the strength of the influence of the outside world exceeds a finite threshold value. We briefly discuss possible tests of this prediction, and its implications.

We discuss an approach to reducing the number of events accepted by anomaly detection systems, based on alternative schemes for interest-ranking. The basic assumption is that regular and periodic usage of a system will yield patterns of events that can be learned by data-mining. Events that deviate from this pattern can then be filtered out and receive special attention. Our approach compares the anomaly detection framework from Cfengine and the EventRank algorithm for the analysis of the event logs. We show that the EventRank algorithm can be used to successfully prune periodic events from real-life data.

Today, little is known about the costs associated with the manageability of a system. This lack of knowledge has a deep impact both on the quality and the performance of managed networks and services and even the management system itself. Thus, it becomes crucial to assess this cost and better understand the performance of management systems both with common metrics and performance evaluation methodologies. Based on empirical observations and analytical results, we identify interesting performance behavior of JMX (Java Management eXtension) based-management frameworks.

Management of pervasive systems cannot rely on human intervention nor centralised decision-making functions due to their complex and intrinsically mobile nature. In previous work, we proposed the concept of a self-managed cell (SMC) as an architectural pattern for building ubiquitous applications. A SMC consists of hardware and software components that form an autonomous administrative domain. SMCs may be realised at different scales, from body-area networks, to an entire room or larger settings. However, to scale to larger systems it is necessary for SMCs to collaborate with each other, to federate or compose in larger SMC structures. We describe here the main abstractions we have defined and explore future directions towards this goal.

In this paper we identify the challenges in enabling self-management for Unmanned Autonomous Vehicles and briefly outline our policy-based approach for realising a self-management framework. We also present preliminary results.

Systems with more and more sophisticated self-management capabilities are introduced in many application domains, such as robotics, the military, aerospace, ground vehicles, IT network and systems management. So far, descriptions of such systems are neither decomposed into common modules nor aligned to a common reference model. For this reason, it is hard to compare them regarding common cross-domain properties of systems with self-management capabilities (SwSMC). It is also hard to transfer design knowledge from one application domain to another. A review of related work reveals few publications with guidance to analytical decomposition into functional modules and the identification of patterns specific to the structure and behavior as well as design and use of self-management capabilities.

To approach this problem, by using cross-domain evaluation criteria and identification of common patterns we can achieve uniformly structured system descriptions. The approach will be implemented by a common system model of SwSMC, an evaluation sheet for such systems, an evaluation results repository, a pattern catalogue and a set of recommendations how to apply the patterns during the design of such systems and keeping an eye on effectiveness and efficiency. So far, a system model and the evaluation sheet have been created, and some evaluations of research prototypes have been carried out. Despite the low number of evaluations so far, first patterns stand out.

In connection with port scan and worm propagation in the Internet, we address in this paper the problem of estimating the ber of destinations communicating with a given source. We propose a computational and memory-efficient technique of finding the top-talker sources. The proposed algorithm is tested against actual data (NetFlow records from the interconnection IP backbone network of France Telecom).

Policy conflict analysis processes based solely on the examination of policy language constructs can not readily discern the semantics associated with the managed system for which the policies are being defined. However, by developing analysis processes that can link the constructs of a policy language to the entities of an information model, we can harness knowledge relating to relationships and associations, constraint information, behavioural specifications codified by finite state machines, and extensive semantic information expressed via ontologies to provide powerful policy analysis processes.

The Internet is based on an X-Domain topology: interconnected domains are managed by independent actors (Fig.1). Deploying critical services (e.g. VPN, video-conference etc.) over such a topology requires to be able to guarantee For this, and to guarantee privacy, , also called , are committed pairwise between domains. While local QoS control issues on each participating domain can be considered solved [4], open problems persist in the field of end-to-end QoS provisioning and monitoring for The key factors to be taken into account are the heterogeneity, independence and privacy requirements of the individual domains.