Quality of Service and Dependable Computer Networks

Sponsors:

Hungarian-Italian Intergovernmental S & T Cooperation Programme

Period:

2004-2007.

Participants:

Universita di Torino, Dipartimento di Informatica;
Universita di Firenze, Dipartimento di Sistemi e Informatica;
Universita di Messina, Istituto di Informatica e Telecomunicazioni;
Budapest University of Technology and Economics, Department of Telecommunications;
Budapest University of Technology and Economics, Department of Measurement and Information Systems.

Project aim:

The most relevant technological revolution in the past few years has been the easy connection of the computer systems of all the world in a single network. Internet has changed the way in which human can communicate each other via computer networks. However, the challenge of the next decades in the computer network area will be to provide integrated services with a preassigned quality of service (QoS). By QoS we mean, in this context, that the user is guaranteed that the communication network to which he is connected will provide him the requested services with a prescribed level of performance (measures according to various possible indices like: total time, response time, information flow rate or throughput (in bit/sec), etc.). To achieve this goal, the hardware structure of the network (links and nodes) must be highly dependable, and the software must be highly reliable.

The research project will be developed over various and interconnected technological fields. Research on the analysis methods and evaluation techniques, experiments, design, both for software and hardware. In particular, design methodologies need to be developed starting from the specification (in some language suitable to designers) to the evaluation phase. This requires different skills to be involved in the project. From the Italian side, the groups participating in the project are mainly involved in the dependability area, and in the evaluation analysis and design of communication networks. The groups of the Hungarian side have experience in formal modelling and verification and in QoS analysis of specific computer and communication systems. Below we give a short introduction to the major research challenges of the field:

• Software specification language: Unified Modeling Language (UML) is considered now a standard for the specification of software systems The project will focus on the assessment of the feasibility of the transformational approach from UML visual models to mathematical models amenable to the automatic analysis of performability and dependability.
  In particular, the project aims at consolidating a systematic method for the analysis of dependability utilizing Stochatic Petri Net (SPN) models. SPN models allows to exploit the information contained in UML-based specification of the system and supports the analyst from the early stages of the software development process.
  The approach to the construction of SPN models is based on net-composition and on a system layered view, to allow mastering the complexity of modeling fault-tolerant systems and to support the reuse of models. An interesting topic of research will be the exploitation of the General Resource Model of the UML to identify resource and service models and their relationships.
  This research activity focuses mostly on the integration of stochastic modeling, solution algorithms of model-based evaluation with software engineering modeling paradigms and languages (like UML), as well as on practical software/hardware architectural and construction issues.
• Network dependability analysis: Evaluation of the quantitative level of network dependability requires the development of suitable analysis methodologies. Since the load on the network is changing, and the accessability of the computers depends on their availability, the models of the network must be based on a stochastic evaluation. During the project we will develop studies and experiences on different methodologies, ranging from classical dependability models like Fault-tree to more sophisticated techniques like Petri nets and Queuing Networks. In particular, one objective of the common effort will be the possibility of integrating in a single modelling framework stochastic and deterministic timing to cope with QoS indices related to real time issues.

Further information (at DMIS):

István Majzik, Ph.D.