Analyzing the effect of imperfect debugging on software fault detection and correction processes via a simulation framework

Abstract

During a debugging operation, there is a high probability that an additional fault will be introduced into the program when removing an existing fault. Thus, perfect debugging is an ideal but impractical assumption when modeling software reliability. If the debugging of a software system is imperfect, more faults may be introduced and detected. In such cases, it may be necessary to add more staff to the debugging team to share the load and ensure the quality of the software. To investigate the effects of imperfect debugging, we simulate the fault detection and correction processes by a single-queue multichannel queuing model with feedback. In this paper, two debugging procedures are discussed. The first, called Procedure_perfect_debugging, is based on a single-queue multichannel queuing model and deals with the case of perfect debugging. Then, we relax the restriction on perfect debugging, and further propose Procedure_imperfect_ debugging based on a queuing model with feedback to address the case of imperfect debugging. We demonstrate the implementation of the procedures via two case studies in which we quantify the effects of imperfect debugging in terms of throughput, time consumption, and debugger utilization. Finally, based on the measurement results, we determine the most suitable staffing level (i.e., the number of debuggers required) for a debugging system under different degrees of imperfect debugging.