In any system, good management depends on access to good information. In real-time systems, however, it is not enough to have access to correct data; that data should also be referenced to time. In distributed systems the validity of data is affected by many factors. Besides the errors that can occur because of hardware or software failures, inconsistency can occur through other factors such as uncontrolled access to distributed data, or because of uncompleted access to duplicated data. Many factors contribute to the incorrectness of data in the domains of value and time. This thesis sets out to investigate the requirements for handling hard real-time data in distributed databases, and suggests an architecture that allows access to such data in a timely and controllable manner. The nature of real-time data, and the requirements for accessing it in the presence of hard real-time and distributed-access constraints, are investigated. Current techniques that maintain the correctness of concurrent access to data in systems were investigated and found to be inappropriate in hard real-time situations. The need for suitable real-time techniques is therefore obvious and urgent, and so an approach that maintains the correctness of distributed access to data under real-time constraints, is proposed here. An experimental attempt to implement and measure the effectiveness of the proposed techniques using an existing environment is described, followed by a proposal for a more complete, fieldbus-based environment. The thesis concludes by showing the results of the work, as well as further research that could be done to obtain solutions to the problems described and partially solved in this thesis.

The steady increase in the use of automation systems in industries such as manufacturing and transportation has created the need for low-cost, robust, real-time communication systems. Various standard communication proposals, such as those based on the MAP and fieldbus initiatives, have been developed to address this need. An important and essential feature of any such real-time communication systems is, it is suggested here, support for clock synchronisation. The need for this arises from the fact that all the data that is read by a real-world controlling process, and then used by a particular node associated with that process, will necessarily be related to the temporal characteristics of the process under control. The Controller Area Network (CAN) fieldbus, initially developed for use in the automotive industry, is finding increased use in the more general industrial automation field, because of its simplicity, robustness and low cost. CAN, however, does not offer direct facilities to support real-time clock synchronisation between the interconnected nodes. Thus, the aim of the work presented in this thesis is to develop techniques that will allow real-time clock synchronisation over the CAN protocol. The solution, which is realised as a hardware-assisted software system, will allow CAN to support truly real-time communication over an existing, commercial system. In this thesis, a study of other possible protocols has been made, leading to the conclusion that CAN is industrially important for use in truly real-time applications. The need for clock synchronisation is reviewed, and possible solutions are surveyed. A hybrid scheme is suggested as being the most promising solution to the problems investigated, in that it balances the performance and cost criteria. Before proceeding to an actual hardware/software design of such a solution, the system was modelled, simulated and analysed. This provided a deep insight into the protocol, helped to resolve many vague manufacturer's specifications, and showed that the proposed solution is viable - especially in terms of the traffic loading introduced. A fully operational 4-layer PCB version of the proposed solution was then developed, and the excellent results achieved are reported here. Based on this, suggestions for full commercialisation have been derived.

A central issue in significantly large industrial distributed computer control systems, but one largely ignored by systems designers, is the question of the real-time handling of data. Recently, though, it has become evident that many installed systems are not able to meet the requirements of full plant-wide integration and synchronization. It is becoming clear that future distributed control systems will only be capable of performing the very exacting tasks asked of them if they are supported by an inherently time-conscious communication system, together with an ability to handle data from a variety of different computer-based devices, in a cohesive fashion. A glimmer of hope appeared with the General Motors-inspired MAP initiative. Although a milestone in providing standardized computer interconnection, MAP does not adequately address the all-important issue of real time. When this issue is, however, fully understood, it is found that it invades all aspects of systems design, where a primary requirement is for the totally deterministic handling of data - both in transmitting the data and in processing it. As a consequence, in the development of an industrial communications system, all supporting components must be carefully designed to ensure completely predictable performance. This thesis tackles the design issue in the particular context of communication protocols. The thesis postulates that it is vital to design protocols which are not only logically incorrect, but also have a fully defined temporal behaviour. The question then becomes one of identifying suitable tools for use in the specification, design and verification of protocols - both logically and in the time-domain.

Over recent years there has been considerable interest in the use of new programming techniques, derived from artificial intelligence (AI) research, in computer control. The reason for considering these techniques is the ever-growing desire for competitiveness in manufacturing and process plants. Two particularly important techniques to have emerged from AI research are expert systems and neural networks. In parallel with the growing need for efficiency, has been the desire to improve the safety and reliability of computer controlled systems. Achieving this requires that any software in the system must have reliable real-time performance - that is, it must produce results which are correct and on time. While progress is being made towards this end in the field of conventional software, the emergence of AI techniques is creating a new set of problems. This thesis examines the reasons why reliable AI systems must be developed, and highlights the differences between AI techniques and more-traditional programming methods. It is argued that true real-time performance can only be obtained from a resource-adequate system and that a state-based approach offers considerable advantages for a real-time expert system. An architecture is proposed which will provide the basis for developing expert systems for hard real-time applications. To overcome the predictability problems normally associated with the inference engine of an expert system, a knowledge-base compiler is presented which produces, from an expert-system ruleset, a procedural function with a known execution time. To validate this approach, a case study is presented in which an expert system has been developed for the control of a aircraft during landing. This example demonstrates the use of a rule-based expert system in a hard real-time environment. An evaluation of the ideas presented is given which identifies some of the limitations of the techniques discussed.

In distributed real-time systems, data are time-dependent and the time notation, which marks the validity of data, is an indispensable component of data. Therefore, setting up a common time base in distributed real-time systems to provide all computers in the system with a consistent view of time inevitably becomes important. Clock synchronization is an approach to establishing such an greed global time base in a system. The aim of work is to investigate clock synchronization in a distributed real-time system. A hybrid synchronization method is adopted in the project. By this means, each computer in the system periodically exchanges time messages with other computers via the existing communication network, and thus knows the clock time of every other computer which participates in the clock synchronization. Then, it applies a synchronization algorithm on the received time data to compute an approximate global time for the system, and finally adjusts its local clock to the computed global time. An approximate global time is thus reached and maintained. In this dissertation, various approaches are studied and compared. A hybrid scheme is proved to be the currently dominant solution to the problem investigated, in that it balances the performance and cost criteria. The fault tolerant averaging algorithm selected for this attempt is discussed and analysed, and its software implementation is described. The design and silicon implementation of the associated VLSI clock synchronization unit device are also introduced. The experiments with clock synchronization on an Ethernet system are presented. The results of the experiments are given and analysed to support the conclusions that the technique developed in this work is feasible and can be used in distributed real-time systems. On this basis, recommendations for further research in clock synchronization are suggested.

No description available.

No description available.

No description available.

Precise operation of real-time systems depends on functionally correct computations that are delivered within imposed timing constraints. These temporal requirements are often modelled and verified assuming a priori knowledge of the Worst-Case Execution Time (WCET) of each task. Due to complexities resolving the actual WCET, estimates normally suffice. These estimates should be safe, so as not to compromise temporal correctness, and accurate, in order to maximise the often limited system resources. The aim of WCET analysis is to therefore compute a WCET estimate that is the actual WCET. To date, the predominant research direction has been static analysis, which builds both program and processor models, and can therefore provide rigourous proofs regarding safety. However, the real-time sector is being infiltrated by more advanced processors that complicate processor modelling sufficiently so that simplfying assumptions are needed. Such assumptions lead to varying degrees of overestimation, depending on processor configuration. On the other hand, current end-to-end testing practices - most often employed in industry - do not target WCET estimation and could therefore underestimate unless the longest path is triggered. This is further complicated by advanced processors as the WCET can depend on a rare sequence of events at the architectural level, and not necessarily on the input causing the greatest number of operations. In this thesis, we combine the relative strengths of testing and static analysis through a Hybrid Measurement-Based (HMB) framework based on a new program model, the Instrumentation Point Graph (IPG). We present an algorithm to construct the IPG from a reducible CFG* - an augmented Control Flow Graph (CFG) - such that arbitrary irreducible IPG loops are identified on the fly. Using these structural properties, we show how to map loop bounds obtained through static analysis onto the IPG and also how to extract observed loop bounds from timing traces. However, since the IPG does not provide a means per se to compute WCET estimates, we remodel two common calculation techniques so that they pertain to arbitrary IPGs. For the purposes of tree-based calculations, we present an algorithm that decomposes the IPG into a new hierarchical form, the Itree; we also present the timing schema used to drive the calculation over the Itree. However, we show that the Itree representation must make a space/precision trade-off when modelling arbitrary irreducible IPGs, ultimately resulting in a margin of overestimation. As a consequence, we rework the Implicit Path Enumeration Technique (IPET) so that it applies to the IPG.

No description available.