Krishna Gopal

Some aspects of reliable software packages

Abstract Very few standard computer programs exist and practically no standards exist to cover common work practices and technical goals, e.g. measurement of software reliability. Success or failure of a software project is mainly determined by the technical skills and the management capabilities of the involved personnel. This paper reviews different aspects of reliable software packages, e.g. software errors, quantitative evaluation of reliability, prevention, detection, diagnosis methods and recovery from malfunctions. The paper also identifies areas in which future research activities should be directed.

A new method for reliability optimization

Abstract A new method for optimizing reliability by use of redundancy in systems which may be complex or series, is described. The system is subject to multiple constraints which need not be linear. The redundant units are allocated to various subsystems of the system in direct proportion to the normalized partial derivative of the system reliability with respect to component reliability of the subsystem. This solution is converted into integer solution by rounding to the nearest integer and an optimal allocation is then obtained by modifying the above integer solution. The method is computationally simple, fast and does not use complicated mathematics. Examples have been included to illustrate the method.

Reliability optimization in systems with many failure modes

Abstract The problem of reliability optimization in redundant systems with two classes of failure modes has been solved using various mathematical programming techniques which are quite involved and slow. In this paper, solution to this problem has been presented using normalized partial derivatives of the system reliability with respect to component reliabilities in each class of failure mode of the subsystems. In this approach the number of redundant components to be used in each subsystem are first determined. These components are then arranged in a suitable configuration so as to give maximum sybsystem reliability. The method is devoid of any involved mathematics and is computationally simple and fast. An example treated by Tillman has been illustrated and it is observed that the solution obtained is identical when subsystem structures are limited to Tillmans form, but an improved solution is obtained when flexible subsystem structure is allowed.

A new approach to reliability optimization in general modular redundant systems

Abstract The paper proposes a new technique for solution of reliability optimization problem in modular redundant systems whose sybsystems may be in series or complex configuration and are subject to multiple linear and/or nonlinear constraints. Modules within a subsystem may be statistically dependent or independent. In each subsystem redundant modules can be added in any one of the redundant modes (active parallel, standby or partial (k-out-of-n: G) redundant mode) or a more reliable module can be used in order to improve the system reliability. The gain in system reliability by allocating all the available resources to one subsystem only are determined for all subsystems and to the subsystem giving the maximum gain one redundant module is added or a more reliable module is used in that subsystem. The method has been illustrated by examples and results compared with existing methods.

Hardware vs software reliability—A comparative study

Abstract High levels of reliability can be predicted and achieved as far as the hardware portions of modern large and complex real-time computer-based control systems are concerned but software is a critical part and plays a vital role in influencing the overall system reliability. This paper presents a comparative study between hardware reliability and software reliability from three definitions: top level, intermediate level and low level based on users point of view, system designers point of view and quantitative measurement point of view, respectively. Also discussed is the hardware reliability theory vs software reliability theory on the basis of analysis of malfunctions and prevailing reliability trends.

A new approach to optimal redundancy allocation for complex networks

Abstract In this paper, a general approach to redundancy optimization in mixed redundant complex networks is presented. The problem of redundancy optimization for complex networks is broken into several simpler and non-interacting smaller problems of optimization for series networks. The best solution of these series problems is taken as the optimal solution for the complex network. A new heuristic criterion is introduced for solving series problems. A redundant component is added to the stage with minimum value of weighted sum of relative increment in resources. This procedure is continued until no more redundant components can be added within the available resources. The method is simple, computerizable, fast and capable of handling networks subject to any number of constraints which may be linear and/or nonlinear. The method has been illustrated by an example and results compared with other methods.

A heuristic method of link reliability assignment for maximal reliability

Abstract Link reliability assignment is the problem of determining the reliability values to be assigned to the various links of a given network from a given set of link reliability values in order to achieve maximum reliability for the given network. Reliability to be maximized may be the s-t reliability or global reliability. The methods available for assigning reliability values to various links require the knowledge of the network reliability function and its several evaluations. The proposed method, which is heuristic in nature, does not require the reliability function and its evaluations; it only needs the network s-t paths (or trees for global reliability). Frequencies of occurrence of various links in paths of different cardinalities (or trees for global reliability) are found. An index, termed the importance index, which is a function of frequency of occurrence, number of paths and cardinalities of paths, is defined and determined for each link. Links are assigned reliability values in the same order as the order of importance index values. The proposed criterion has been tried on several networks. The network reliability obtained is maximal.

An algorithm for determining the most reliable path of a network

Abstract This paper present an algorithm for determining the most reliable path between two specified vertices of a communication network, taking into account both the reliabilities of nodes and links. The proposed algorithm is based on the basic idea of node removal technique, which is generally used for finding Boolean transmission functions between input-output terminals in the analysis of bilateral switching networks. The proposed algorithm can be applied to oriented or non-oriented, symmetric or non-symmetric networks and is capable of handling networks with hundreds of nodes and links. The proposed algorithm is a good time saver and efficient for large networks, as compared to other existing methods.

An improved method of selecting network topology for optimal terminal reliability

Abstract Given the permissible network topology and cost, it is generally required to determine the topology so that source-to-terminal reliability is maximum. Such an exercise may be needed for any type of network; like that of computers, communication lines, transmission lines, pipelines etc. In this paper two heuristic methods of determining such an optimal network are discussed and compared with that of Chopra et al. Microelectron. Reliab.24, 911–913, 1985. It is shown that the method of Chopra et al. fails quite often and the ones proposed are found to be successful in majority of cases.

Detection and diagnosis of software malfunctions

Abstract Today, providing computer software involves greater cost and risk than providing computer equipment, because hardware is mass produced by industry using proven technology, while software is still produced mostly by the craft of individual computer programmers. Software reliability improvement is achieved through not only structure and care in design, implementation and verification of software, but also through effective use of redundancy in the form of robust data structures and information about what constitutes expected behaviour of software. Quick detection of malfunctions minimizes the damage caused by malfunction and leads to a rapid recovery. This paper summarizes techniques and tools used for detection and diagnosis of malfunctions occuring in software systems.