Proposal skelteon

13 Network and Distributed System Management

Network and Distributed Systems management is a broad discipline covering all the adaptation and control activities performed on a system after it is deployed to ensure its correct and optimal functioning. In addition, Network and Systems management also covers activities such as provisioning and capacity planning which are performed during the deployment phase. Research in network and systems management can be broadly subdivided in two areas: (i) the design of systems management infrastructures which includes the development of new techniques, tools and architectures for systems management and (ii) the application of those techniques to various domain specific areas such as network quality of service (QoS), multimedia, and cellular telecommunications.

However, to be able to manage a system it is necessary to monitor its behaviour and to perform control actions which adapt its behaviour and functioning. Thus, although partial monitoring of a system can be achieved non-intrusively, monitoring and especially control activities can never be entirely separated from the design of the system itself. Thus, many of the projects focussing on network and systems design invariably comprise and use innovations in management techniques for making them adaptable and re-configurable. In particular, many traditional management services and techniques such as instrumentation, event handling, management by delegation and policy-based management are being considered as important components of the design of mobile and pervasive infrastructures which rely on asynchronous event notifications to adapt their functioning. Therefore, many of the projects mentioned in this chapter overlap with those in other chapters. In particular, projects on network QoS are also addressed under network design and multimedia (see Chapter 11), and projects on pervasive and mobile systems often encompass adaptation techniques and security projects which cover security management in their scope.

13.1 Current and Recent Work

There are continuing research efforts towards providing guaranteed Quality of Service and reliable communications in multiservice networks. Current investigations focus on Traffic Engineering (including load measurement and modelling) [Wolfinger and Kühn 2002][Wolfinger et al 2002] policy-based management, and the provision of adaptive and programmable networks. Policy-based management is seen as an important technique towards providing uniform control of network resources and driving the adaptation of QoS mechanisms according to changes in application requirements, variations in network usage or in response to failures and performance degradations [Lymberopoulos et al 2003]. Advances have been made towards policy specification [Damianou et al 2001], policy deployment [Dulay et al 2001], the implementation of tools for policy-based systems [Damianou et al 2002] and policy analysis and refinement [Bandara et al 2003]. The ANDROID project has been using policy-based management of proxylets which execute in the application layer within servers in the network and hence provide an Application Level Active Network. Application level adaptation techniques such as adaptive video encoding and information dispersal have also been investigated [Richter 2001].

In parallel with the work on network QoS, substantial effort is being devoted towards the development of reliable, fault tolerant architectures and control mechanisms for a variety of application areas including the automotive industry. A significant theme across many of the ongoing studies is the development of new techniques and infrastructures capable of providing adaptive behaviour at both the application and the network level in large distributed systems. This work is becoming even more critical in mobile and ubiquitous systems which are characterised by numerous dynamic changes in network connectivity, usage context and service availability.

13.2 Future Trends

Work towards providing QoS in multiservice networks will certainly continue although there is a need for better integration between the network provisioning, traffic engineering, and management systems.

Increasing emphasis will be placed on the development of mobile and ubiquitous computing environments (see Chapter 5). Entities in these environments must exhibit a high degree of “intelligence” by adapting both their requirements and their behaviour to frequent changes in their environment and the context in which they evolve. Frequently, entities in these environments such as laptops, PDAs or embedded computers have limited resources and limited network communication capabilities. Management in such environments is a significant challenge as the need for adaptability, programmability and context awareness is exacerbated. To address these issues new techniques for providing dynamically adaptable control in embedded devices are necessary and highly dynamic management frameworks which combine a variety of adaptation mechanisms and techniques must be defined. The scale and widespread deployment of such systems will require new self-managing devices which can collaborate within the context of self-managed and dynamic coalitions. A new facet in ubiquitous systems is the need for context-aware management and management of context services [Hegering et al 2003].

The main challenge in the years to come is therefore to identify common architectural patterns and techniques for self-management that can be extensible and scale down to dynamic coalitions of embedded devices, e.g., for body area networks and scale up to large enterprise systems.

Increasing security concerns have emphasised the need for environments where security is not the result of a static configuration manually changed by a system administrator but where the security system continuously adapts and reacts to events such as failures, intrusions and interactions with other systems. This requires a better integration between the management and the security systems and renewed efforts towards adaptive security management.