一般注記Information technology (IT) systems have become essential to our society as they are extensively used in our daily lives, economic activities, and social infrastructures. The effectiveness of software services delivered through IT systems directly impacts on human lives, organizational productivity, and social safety. Since malfunctions of such IT systems can cause serious problems, dependability and performability design of these systems is crucial. A fundamental challenge is the difficulties in dealing with various uncertainties in complex software systems, such as demand changes and component failures, that are inevitable during their operations. In this thesis, the applications of stochastic models are presented, particularly for dependability and performability system design and operation. Stochastic models are used to capture dynamic failure-recovery behaviors of systems as well as fault-tolerant configurations. Since recent IT systems are becoming more software- intensive owing to increased computing power, the impact of software faults and errors must be taken into account. In this thesis, dependability and performability design of software- based IT systems are discussed in three areas. In the first area, software aging problem, which is often observed in long-running software systems, is discussed. Against software aging, the effective countermeasures, software rejuvenation, and software life-extension, are evaluated. In the second area, the effectiveness of storage array configurations and their data management operations are evaluated. To achieve high data-availability with acceptable performance, optimal storage array configurations and data backup operations are analyzed. In the third area, the issue of cloud resource management to avoid performance problems due to resource contention is discussed. Virtual machine reallocation in a cloud data center and additional physical server procurement are considered. In all three cases, dynamic behaviors of systems are captured by stochastic models that are used to quantitatively evaluate the performance measures by using analytical solution techniques. Quantitative assessment enables design improvement and maintenance optimization toward achieving highly available and high-performance software-based systems.
identifier:oai:t2r2.star.titech.ac.jp:50404618
一次資料へのリンクURLhttp://t2r2.star.titech.ac.jp/rrws/file/CTT100759679/ATD100000413/thesis_15D37010_町田文雄.pdf (fulltext)
連携機関・データベース国立情報学研究所 : 学術機関リポジトリデータベース(IRDB)(機関リポジトリ)