SREPT: software reliability estimation and prediction tool

Abstract

Several tools have been developed for the estimation of software reliability. However, they are highly specialized in the approaches they implement and the particular phase of the software life-cycle in which they are applicable. There is an increasing need for a tool that can be used to track the quality of a software product during the software life-cycle, right from the architectural phase all the way up to the operational phase of the software. Also the conventional techniques for software reliability evaluation, which treat the software as a monolithic entity, are inadequate to assess the reliability of heterogeneous systems, which consist of a large number of globally distributed components. Architecture-based approaches are essential to assess the reliability and performance of such systems. This paper presents the high-level design of a software reliability estimation and prediction tool (SREPT), that offers a unified framework consisting of techniques (including the architecture-based approach) to assist in the evaluation of software reliability during all phases of the software life-cycle.

Introduction

Software is an integral part of many critical and non-critical applications, and virtually any industry is dependent on computers for their basic functioning. As computer software permeates our modern society, and will continue to do so at an increasing pace in the future, the assurance of its quality becomes an issue of critical concern. Techniques to measure and ensure reliability of hardware have seen rapid advances, leaving software as the bottleneck in achieving overall system reliability.

Software reliability is defined as the probability of failure-free software operation for a specified period of time in a specified environment [43]. Its evaluation includes two types of activities [32]: reliability estimation and reliability prediction.

• Estimation. This determines the current software reliability [9] by applying statistical inference techniques to failure data obtained during system test or during system operation. This is a measure of “achieved” reliability, and determines whether a reliability growth model is a good fit in retrospect.

• Prediction. This determines future software reliability based upon a fitted reliability growth model or other available software metrics and measures.

Various approaches have been proposed to achieve reliable software systems and these may be classified as [26]

• Fault prevention. To avoid, by construction, fault occurrences.

• Fault removal. To detect, by verification and validation, the existence of faults and eliminate them.

• Fault tolerance. To provide, by redundancy, service complying with the specification in spite of faults having occurred or occurring.

• Fault/failure forecasting. To estimate, by evaluation, the presence of faults and the occurrence and consequences of failures.

Of these, fault/failure forecasting techniques have received widespread attention with the development of software reliability growth models [6]. These models capture the process of fault detection and removal during the testing phase in order to forecast failure occurrences and hence software reliability during the operational phase. Many of these models have been encapsulated into tools [32] like SMERFS, AT&T SRE Toolkit, SoRel, CASRE [31], RAT [28], M-élopée [45]. The problem with applying such tools effectively towards improving the quality of software is that they need data about the failures detected during the testing phase in order to estimate software reliability model parameters. Thus they assess reliability very late in the life-cycle of the software, and it may be too costly to take remedial actions if the software does not meet the desired reliability objective. Also, these tools do not take into account the architecture of the software. Though some techniques have been proposed [22], [25], [36], [40] to perform architecture-based software reliability prediction, there is no special purpose tool available yet. Thus, there is a need for a tool that can track the quality of a software product and provide insights throughout the life-cycle of the software. In this paper we present the high level design of a tool which provides a unified framework encompassing several techniques to monitor the reliability of a software product during the entire software development process.

The rest of the paper is organized as follows: in Section 2 we provide the motivation for a new tool. Section 3 presents the high-level design of SREPT. Section 4 provides an illustration of the use of SREPT in estimating software reliability. In Section 5 we comment on SREPT’s expected impact in the area of software reliability estimation and prediction. Section 6 concludes the paper by summarizing the features of SREPT that are presented in this paper.