E.M. Scharf

Assessing multiple choice question (MCQ) tests - a mathematical perspective

The reasoning behind popular methods for analysing the raw data generated by multiple choice question (MCQ) tests is not always appreciated, occasionally with disastrous results.This article discusses and analyses three options for processing the raw data produced by MCQ tests. The article shows that one extreme option is not to penalize a student for wrong answers or for missing out questions, and the other extreme option is actually to penalize both aspects.The intermediate option of focusing on the number of questions actually attempted while penalizing wrong answers can be regarded as the fairest. In this case blind guessing will on average not help the student, although partial knowledge will lessen the negative impact on the final overall score. There are still many interesting challenges in designing techniques for MCQ tests.

A fault diagnosis system for the connected home

The connected home of the future, in which all consumer appliances in a home are networked together, is close to becoming the connected home of today. This article explores the role of fault diagnosis in such an environment and explains how agent technology may be applied. The article outlines the need for future standards that can maximize the benefit of a shared diagnostic system.

Tight bounds for the tail of the packet waiting time distribution in buffered networks

In this paper we consider the end to end delay through a series of queues in which the service time is fixed and equal to the transmission time of constant length data packets at constant bitrate, for example real time delay sensitive traffic in IP or ATM networks. In these cases, algorithms are required either for on-line admission control or for planning and provisioning of service partitions within such networks. Here we develop concise, accurate and easy to use solutions for the end to end delay through this series of queues, and provide numerical results quantifying the high degree of accuracy obtained. Additionally we compare against a Chernoff Bound approach, showing that our solutions are more accurate. Copyright

Model based risk management of security critical systems

This paper describes a novel framework for a risk management process involving a model-based approach, developed as the main objective of CORAS (IST-200

Customer choice in a multi-service residential access network environment

TORRENT is an EU-Supported Framework 5 project that is showing how intelligent control and negotiation capabilities can be built into access networks so as to give the residential customer the ability to choose among a variety of service offerings, in order to achieve an optimal combination of service parameters. This access should be transparent to the type of the core network being addressed. The realisation of a prototype service for a resource management system based on intelligent agent technology is now underway.

ATM Network Simulation Support for TMN Systems

RACE Project R2059 ICM is developing cell rate ATM network simulators for use in the development and testing of a generic Telecommunications Management Network (TMN) test bed. The simulator is used as a substitute for a physical ATM network and enables the TMN to be tested in a controlled environment before interfacing with the real network. The simulator is able to provide a number of major enhancements over current ATM networks enabling more extensive and thorough evaluation of the TMN. This paper describes the functionality of the simulator and the underlying modelling principles. It emphasises the interaction of the simulator with the TMN system and the facilities of the simulation models to aid this interaction. The paper shows how the simulator is being used to model a Virtual Private Network for the final ICM case study, in particular concentrating on the use of generic reusable components (User, Call Handler and Switch models) for simulating additional entities such as the Customer Premises Network, the Interworking Unit, and the Gateway between two ATM networks.

A priority queue with superimposed geometric batch arrivals

In this paper, algorithms are derived that permit the exact evaluation of the steady state waiting and system time probabilities, for two different priority customer types, in a discrete time, batch arrival queueing system, which has infinite waiting space. All customers have integer, but non-zero, service times that are generally distributed. Both preemptive and non-preemptive scheduling is considered. Numerical results are presented indicating how these algorithms can be used to study certain problems arising in future telecommunications networks.

A new charging scheme for ATM based on QoS

This paper describes a new approach to charging for ATM called the ‘quality of service (QoS)-based charging scheme’. In this scheme, traffic resources are distributed among buffers established to support combinations of ATM transfer capabilities and qualities of service. The buffers are dimensioned according to M/D/1/K and ND/D/1 queuing analysis to determine the buffer efficiency and quality of service requirements. This dimensioning provides the basis for fixing the price per unit of resource and time. The actual resource used by a connection is based on the volume of cells transmitted or peak cell rate allocation in combination with traffic shapers if appropriate. Shapers are also dimensioned using the quality of service parameters. Since the buffer efficiency is dependent on the quality of service requirements, customers of ATM networks buy quality of service. The actual price of a connection is also related to the amount of the resource purchased as well as the time of the day at which a connection is made, and the geographical location of the destination switch. The QoS-based charging scheme meets the requirements of customers and of network operators. Its performance compares very favourably with that of a number of well-known existing ATM charging schemes. Copyright