Petre Dini

Architectural design of adaptive distributed multimedia systems

In the context of a collaborative research project funded by the Canadian Institute for Telecommunications Research (CITR), we have developed a prototype system for remote access to News-on-Demand. This system allows the user to remotely access a multimedia database, containing news clips in the form of multimedia documents, over ATM and other types of networks. Special attention is given to quality of service (QoS) negotiation and adaptation. For instance, a given document may exist in different versions on different sites and possibly corresponding to different presentation qualities, such as video and audio quality, size of display and cast. A graphical interface is available for the user to select his preferences and provides the possibility of obtaining examples of specific quality features. The QoS negotiation and adaptation features allow for the selection of the best configuration for a given user request and for automatic adaptation in case of changes to the system parameters, such as network or server congestion. We present in this paper an abstract architectural design of our adaptive distributed multimedia system for remote access to multimedia databases. This design focused on the aspects of the system which are essential for QoS negotiation and adaptation, which is our main concern in the ongoing CITR research project. The paper gives a functional overview of the system and details its structural and behavioral aspects. An abstract application programming interface (API) is given and issues related to the transition from the abstract system design to an implementation are discussed.

Performance evaluation for distributed system components

The performance evaluation of hardware and software system components is based on statistics that are long views on the behavior of these components. Since system resources may have unexpected behavior, relevant current information becomes useful in the management process of these systems, especially for data gathering, reconfiguration, and fault detection activities. Actually, there are few criteria to property evaluate the current availability of component services within distributed systems. Hence, the management system can not realistically select the most suitable decision for configuration. In this paper, we present a proposal for a continuous evaluation of component behaviour related to state changes. This model is further extended by considering different categories of events concerning the degradation of the operational state or usage state. Our proposals are based on the possibility of computing at the component level, the current availability of this component by continuous evaluation. we introduce a several current availability features and propose formula to compute them. Other events concerning a managed object are classified as warning, critical or outstanding, which leads to a more accurate operational view on a component. Several counter-based events are thresholded to improve predictable reconfiguration decisions concerning the usability of a component. The main goal is to offer to the management system current relevant information which can be used within management policies the flexible polling frequency tuned with respect to the current evaluation, or particular aspects related to dynamic tests within distributed systems. Implementation issues with respect to the standard recommendations within distributed systems are presented. Finally we describe how the reconfiguration management systems can use these features in order to monitor, predict, improve the existing configuration, or accommodate the polling frequency according to several simple criteria.

Agent based management of distributed systems with variable polling frequency policies

Management activities are based on the state of distributed system components, relations of these components, and their behaviour. Since management policies are applied across an abstraction of distributed systems, the quality of decisions is dependent on the representation fidelity of the real system state. Obviously, the data collection process updating the abstract representation of real distributed system components has a major impact on the quality of management decisions. Gathering most topical management data improves the quality of management decisions, but requires a high degree of monitoring activity. This is contradictory to the request for low impact management systems, where the amount of system resources used for management purpose should be as small as possible.

Towards automatic trading of QoS parameters in multimedia distributed applications

Several existing and emerging studies exhibit demands for the definition, measuring, testing and enhancement of QoS. Since real-time multimedia applications require guaranteed performance services or graceful degradation services many problems are particularly complex. ODP considers QoS aspects embedded within the trading function. Emerging systems portray particular service parameters, which must be considered by an ODP trader. The advent of multimedia leads to new functional services (guaranteed, predictive), new management services (QoS management, synchronization management), new protocols (negotiation, re-negotiation), and new mechanisms (parameter translation, parameter matching). This paper deals with the matching function and particular trading aspects in multimedia distributed applications. In order to offer a more flexible matching and reduce the negotiation dialog, we propose a formal model for service parameters, defining special parameter fields. Each new feature is analysed across multimedia examples, its particularities, its value space, and the customer-provider issues related to it. Based on this model, a precise definition of the matching function which is useful to automate the negotiation of QoS is proposed. An algorithm implementing this function is analysed. A model for the evaluation of contractual QoS at the parameter and service level, and specific formulas are presented.

Formal and Informal in Balanced System Specifications

The paper presents the relation between balanced system specification by a formal or informal documentation. The text is a technical description of experienced variants, rather than pretentious advices on this topics. We claim that the advent of the object-oriented approach, combined with a mix, formal and informal documentation, in the context of automation, represents a possible solution for either the easiness of knowledge transfer, the use of advanced tools for the system design, or a rapid adaptation from large size enterprises to small ones because of the modularity and reuse. Our report focuses first, on existing techniques, object-oriented methodologies, and second, on lessons learned and retained with respect to the system documentation.

Configuration in the framework of open distributed processing

Open distributed processing (ODP) is an ongoing standardization activity by both ISO and CCITT providing an object oriented reference model (RM ODP) for open distributed systems. Certain abstraction levels called viewpoints are introduced by the RM ODP. The formal description technique LOTOS is used to define formal models for the computation and engineering viewpoint by sublanguages LOTOScomp and LOTOSeng. The use of these sublanguages as configuration languages on different abstraction levels is shown and illustrated by an example. The mappability of LOTOS specification in particular from the engineering viewpoint into implementations is outlined for the object space.<<ETX>>