Selecting Optimal COTS Products Considering Cost and Failure Rate

A reliable and cost effective system is crucial in satisfying the requirements of users and managers alike. One way to develop such a system is to use existing COTS (Commercial Off-The-Shelf) software products. As an appropriate process model for a software development project, Boehm [2] recommended the use of COTS products in all possible applications. In the US, COTS products are mandated by several large contractors such as the Department of Defence's Message System and NASA's X2000 technology for deep-space mission [12]. The use of COTS products is also viewed as a way to reduce schedule delay and cost overrun in software development projects. This study addresses an optimization model to develop a modular software system. The system is comprised of a set of serially-executed modules, where each module is configured with only one COTS product among several alternatives subject to an overall system failure rate. The objective function of the model minimizes a system development cost (initial cost). The software system configuration of our model is the same as based on the first reliability model of Berman et. al [1], where the initial cost and reliability were only considered. In contrast, our model includes the system development cost and a system failure rate. Kubat [1] presented a stochastic model to minimize a cost subject to an overall system failure rate. The overall system failure rate of our model is derived with a similar approach in Kubat. Our model does not considered the inter-module failure rate addressed by Littlewood [10], where he stated the inter-module failures as ``Unfortunately, there seems to be little information available about interfacing failures...''. Helander et. al [5] introduced two models (a reliability constraint cost-minimization and a budget constraint reliability maximization) for distributing development effort among software components. Jung [7] presented an optimization model to optimize the life-time costs, i.e. the sum of initial and risk cost. His risk cost is similar to the risk cost introduced by Pham and Zhang [11]. Our model is an integer problem and can be utilized in various phases of system development to manage cost and system failure rate. The following section derives an optimization model and final remarks are presented at the end of this paper. 2. Optimization Model