運用快捷排隊理論方法於軟體開發人員配置與管理之研究

A Study of Applying Express Queueing Theoretic Approach to Personal Allocation and Management in Software Development

張書誠(Shu-Cheng Chang)

指導教授 ： 黃慶育

國立清華大學/電機資訊學院/資訊工程學系/碩士(2014年)

摘要

在軟體開發的過程中，移除軟體錯誤是一項高成本且耗時的任務。因此，對專案管理者而言，除錯過程的評估與預測很重要。在過去，專家學者使用軟體可靠度成長模型來描述軟體錯誤的移除過程。他們假設錯誤偵測是非齊次卜瓦松過程，並且為了簡化模型，大部分的研究忽略錯誤偵測到移除之間的時間延遲。近年來，許多學者意識到除錯所需的時間不應該被忽略，因此將排隊理論應用至軟體除錯的研究。除錯人員需要時間來尋找軟體出錯的原因，並移除錯誤的根源。在考慮除錯時間的情況下，許多學者提出無限伺服器排隊模型來評估、分析軟體專案。有一部分的研究甚至考慮除錯人員的數目限制，進而提出有限伺服器排隊模型。然而，過去的這些研究僅關注在除錯的時間延遲與除錯人員的數量。關於軟體錯誤分類的研究很少。若能將軟體錯誤進行分類，則錯誤將能交派於特定的除錯人員並得到更適當的處理。本篇論文的目的是應用快捷排隊理論方法來降低軟體錯誤的平均修復時間，並協助管理軟體專案，如人力配置、成本控制、資源使用率等議題。軟體除錯的流程可以套用排隊理論，而偵測到的軟體錯誤會進行分類並分配至快捷佇列或是一般佇列。實驗結果顯示，使用此方法可以提升軟體除錯系統的效能。此外，我們開發的模擬程序不僅驗證快捷排隊模型的結果，也提供專案管理者一個實用的工具以仔細監控專案變化。

關鍵字

排隊理論 ； 軟體除錯 ； 專案管理 ； 快捷佇列 ； 軟體錯誤分類

並列摘要

Fault correction is costly and time-consuming in the software development process [1-5]. Thus the measurement and prediction of the fault correction process is important for the software project manager. In the past, researchers used software reliability growth models to describe the fault correction process. They assumed that fault detection is a non-homogeneous Poisson process, and many studies have ignored the time lag between fault detection and fault correction to simplify the models. Recently, the queueing theory has been introduced to the software debugging process, since many researchers are aware that the fault correction time should not be ignored. The engineers need time to identify the reasons for the failures and remove the root causes. Considering the correction time, several infinite server queueing models are proposed to measure and analyze the software project. A small number of research has even considered the limited number of debuggers and has presented finite server queueing models. However, the previous work focused only on the debugging time lag and the number of servers. There is little research about fault classification. If the faults would be classified, the fault can be assigned to a specific engineer to get better service. The aim of this study is to apply the express queueing theoretic approach for reducing the mean resolution time of faults and to support the management of the software project, such as staffing, cost, utilization, etc. The debugging process is modeled by the queueing theory, and the detected faults will be classified and dispatched into either express queue or regular queue. The experiments show that the performance of the debugging system improved while using the proposed approach. Besides, the simulation procedures we developed not only verify the results of the express queueing model, but also provide a useful tool for project managers to monitor the change of the system in more detail.

並列關鍵字

Queueing Theory ； Software Debugging ； Project Management ； Express Queue ； Fault Classification

參考文獻

[4] V. Basili, et al., “Understanding and Predicting the Process of Software Maintenance Releases,” Proceedings of the 18th International Conference on Software Engineering, Vol. 464, No. 474, pp. 25-29, March 1996.

[6] S. S. Gokhale and R. E. Mullen, “Queuing models for field defect resolution process,” Proceedings of ISSRE’06 17th International Symposium on Software Reliability Engineering, IEEE, pp. 353-362, November 2006.

[7] G. Antoniol, A. Cimitile, G. A. Di Lucca, and M. Di Penta, “Assessing staffing needs for a software maintenance project through queuing simulation,” IEEE Transactions on Software Engineering, Vol. 30, No. 1, pp. 43-58, 2004.

[9] M. I. Kellner, R. J. Madachy, and D. M. Raffo, “Software process simulation modeling: why? what? how?,” Journal of Systems and Software, Vol. 46, No. 2, pp. 91-105, 1999.

[11] J. Jacod, “Two Dependent Poisson Processes Whose Sum Is Still a Poisson Process,” Journal of Applied Probability, Vol. 12, No. 1, pp. 170-172, March 1975.