Fault trees have been used for software safety analysis in various safety critical systems. The PRIORITY-AND gate was proposed because the conventional AND gate cannot be used to represent the sequential order of the events. The paper shows that even PRIORITY-AND gate is not expressive enough to represent the relative temporal order of the events precisely. We extend the fault trees with an AND-THEN gate that is the corresponding gate of the logical connective TAND. This increases the expressive power of the fault trees. The AND-THEN gate can represent relative temporal relations precisely.

Software Requirements Specifications (SRS) have been used to fill the communication gap between systems analysts and the end-users. SRSs should satisfy the needs of both systems analysts and end-users. Non-technical end-users require intelligible SRSs while systems analysts need more precise, clear and concise SRSs. Object-oriented methods cannot represent temporal relations between events precisely. However, object-oriented principles are widely used in systems analysis and designing. Hence, there is a need for a software requirements specification language which supports object-oriented analysis methods, represents temporal knowledge precisely and whose representation scheme resembles natural languages. The specification language presented in this paper, GSL, is designed to meet the above requirements. The language is based on First-order Temporal Logic (FTL), which has temporal operators in addition to classical logical connectives and quantifiers. Since FTL cannot represent relative temporal knowledge and it inherits problems with point-based time models, a new logical connective TAND and redefined AND connective are used to represent relative temporal knowledge and to solve the problems with FTL. The language employs object-oriented principles: events, conditions, rules and activities can be represented as objects as well as attributes of an object. However, systems analysts can decide whether to use object-oriented conceptual modeling or not. © 1998 John Wiley & Sons, Ltd.

A way to improve the reliability of the name resolution mechanism in distributed operating systems is proposed. It is shown that the name resolution mechanism can be made highly reliable by achieving a high degree of reliability of the name resolution operation for a limited number of pairs. To take care of the changing reliability needs of a particular user for different objects with the passage of time, the system should provide the flexibility to dynamically change the degree of reliability of the name resolution operation for certain pairs. Implementation examples are presented.>

An efficient method of consistency control of replicated directories is presented. By taking advantage of special characteristics of directories, the method achieves fast access to directories and high concurrency in updating directory replicas. The algorithm differs from conventional mechanisms for concurrency control of replicated data in two aspects: It does not use global locks or global timestamp orderings. Updating operations can proceed without being in synchronization. The algorithm can survive both node failure and network failure. The directory problem, design objectives and related works are described. The system model and consistency control requirements are defined, and the data structures and algorithm are presented. The fault tolerance and recovery mechanism of the approach are discussed, as is the applicability of the algorithm. The approach is evaluated and compared with other works. The detailed algorithm and consistency proof are given.>

A description is given of a novel concept and mechanism, called an object group, that realizes hierarchical, integrated object processing and management. Object groups can be used as a software basis for constructing object-based distributed systems. They support hierarchical control of objects, group-oriented communications, group access control, group resource management, and some generic operations on groups, providing consistent and uniform interfaces. Applications include job control, parallel processing, servers, object pools, hierarchical resource management, and so on. Some design issues are presented, and an implementation on a network-transparent global distributed system, which is under development, is discussed.>

This paper describes firmware and hardware support necessary for constructing easy-to-understand and high performance operating systems including language translators and interpreters. Basic principles are one-to-one correspondence between logical hierarchy and physical hierarchy, and vertical migration. Implementation of monitors in firmware and architectural support for it are discussed, and a sample system is shown. Architectural support for user microprogramming is then discussed and an example is shown. After a total system firmware structure is discussed, an experiment of vertical migration is described. It is shown that a proper selection of modules for migration is extremely important. It is suggested that the direction shown in this paper is one of future directions of computer systems.

An APL interpreter is analyzed to examine how microprogramming can improve its performance. Also examined is how the modularization method or program structure affects performance improvement. Two basic modularization methods, function and data modularizations, are investigated. We rind that the performance gain may reach 100-time speed-up and that a proper selection of modules is very important to obtain the maximum performance gain under a limited microprogram memory. We also find that both the function and the data modularizations provide the same degree of performance improvement despite the finding that they tend to affect the complementary parts of a program.

The Experimental Polyprocessor System (EPOS) is described. It is a new computer system based on the concept of the polyprocessor and is intended to obtain a high degree of extensibility, cost/performance, adaptability and reliability. EPOS is composed of computing modules, input/output modules, and shared memory modules which are connected by multiple independent common buses. Modules and system buses can be added or deleted freely to provide a high degree of extensibility. The computing module is using a 7000-gate LSI computing element PULCE and provides extensive support for user and dynamic microprogramming. The input/output modules and the computing modules are connected equally to the system buses, and the same addressing mechanism is applied to them.

EPOS's operating system structure, firmware structure, functional distribution, and restructuring are described. EPOS processors are functionally specialized to user functions as well as operating system functions. The performance improvement by functional specialization is generally expected to be two-to-ten times. The objective of job scheduling is to maximize the total system performance by assigning each job to the best suited processor. For this purpose and for easy maintenance, jobs can be reassigned dynamically. Graceful degradation is also mentioned.

A system consisting of two identical computer systems connected by a communication line is analyzed by using queuemg theory to obtam the turnaround time d~stribution, the trafftc mtenstty between the two computer systems, and the mtertransfer time distribuUon The analysis ts earned out under the assumptions of Potsson arrivals and exponentml servtce times The queue discipline analyzed ts first-comefirst-served with queue adjustments on both job arrivals and completton~ The problems are formulated as partial difference equations and explictt solutions are obtained.The lesults mdtcate that the turnaround umes are more dispersed m networks than m centrafized systems Another queue dtsctphne, still first-come-first-served but with queue adjustments only on job complettons, ~s analyzed for the trafftc mtenstty between the computer systems.The result shows that the queue dtsctpline wah queue adjustments only on job ¢omplettons produces much less traffic for a htgh input traffic rate than the one wtth queue adjustments on both job arnvals and complettons.Stmulatlon results suggest that the above observation holds independent of the form of mterarnval and servtce ttme dtstnbuttons.