OKAMURA Hiroyuki Hiroshima Univ., Grad.School of Eng., Associate Professor, 大学院・工学研究科, 助教授 (10311812)
DOHI Tadashi Hiroshima Univ., Grad.School of Eng., Professor, 大学院・工学研究科, 教授 (00243600)
OSAKI Shunji Nanzan Univ., Faculty of Mathematical Informatics, Professor, 数理情報学部, 教授 (10034399)
TAKAHASHI Yoshitaka Waseda Univ., Faculty of Commerce, Associate Professor, 商学部, 助教授 (20329064)
KAIO Naoto Hiroshima Shudo Univ., Faculty of Economic Informatics, Professor, 経済科学部, 教授 (80148741)
|Budget Amount *help
¥7,500,000 (Direct Cost : ¥7,500,000)
Fiscal Year 2003 : ¥2,900,000 (Direct Cost : ¥2,900,000)
Fiscal Year 2002 : ¥2,500,000 (Direct Cost : ¥2,500,000)
Fiscal Year 2001 : ¥2,100,000 (Direct Cost : ¥2,100,000)
This research deals with the design and control of computer and communications systems with failures and costs. The purpose is to develop new performance/reliability evaluation methodology (PRET) for these systems by uniting performance evaluation and reliability analysis. During the research period, we obtain the following results :
1.Studies on Diffusion Models and Their Applications
For a class of queues formulated as a generalization of continuous-time Markov processes, Kimura developed a unifying diffusion model that gives an approximate queue-length distribution, being consistent with exact solutions for specific cases. His research results are summarized in three papers, see [3,11,23] in the report. Using a diffusion model with an elementary-return boundary at the origin, Takahashi analyzed an M/G/1 queue with modified services at start-up. As applications of diffusion models to computer/communications systems, Kawanishi and Takahashi  analyzed a multi-address call extraction s
erver in a facsimile communication network, and also Okamura and Dohi analyzed a multi-stage production/inventory system. In addition, Kimura wrote a survey paper  on diffusion models for computer/communications systems.
2.Studies on Software Rejuvenation Scheduling
Software rejuvenation is a preventive scheme for handling transient software failures by stopping the running software occasionally, clearing its internal state and restarting it. Using discrete-or continuous-time stochastic models, Okamura, Dohi and Kato derived analytically the optimal rejuvenation schedule that maximizes, e.g., cost-effectiveness in the steady state ; see [1,7,8,9,14,16,17,18].
3.Studies on Software Reliability Growth Models
The time evolution of reliability growth in software testing can be modeled by a stochastic process called the software reliability growth model Okamura, Dohi and Osaki derived accurate solutions for the problems of analyzing a growth model with modular structure, estimating model parameters, finding an optimal software release time and so on ; see [4,5,6,13,15,21].
Okamura, Dohi and Osaki [10,19,22] developed renewal-process models to obtain an optimal solution of the auto-sleep scheduling problem in computer systems, and Dohi, Kaio and Osaki  proposed a nonparametric method to estimate an optimal repair-time limit in a replacement problem with imperfect repair. Less