Stochastic Modeling of Modern Storage Systems

Thumbnail Image




Xia, Ruofan


Trivedi, Kishor S

Journal Title

Journal ISSN

Volume Title

Repository Usage Stats



Storage systems play a vital part in modern IT systems. As the volume of data grows explosively and greater requirement on storage performance and reliability is put forward, effective and efficient design and operation of storage systems become increasingly complicated.

Such efforts would benefit significantly from the availability of quantitative analysis techniques that facilitate comparison of different system designs and configurations and provide projection of system behavior under potential operational scenarios. The techniques should be able to capture the system details that are relevant to the system measures of interest with adequate accuracy, and they should allow efficient solution so that they can be employed for multiple scenarios and for dynamic system reconfiguration.

This dissertation develops a set of quantitative analysis methods for modern storage systems using stochastic modeling techniques. The presented models cover several of the most prevalent storage technologies, including RAID, cloud storage and replicated storage, and investigate some major issues in modern storage systems, such as storage capacity planning, provisioning and backup planning. Quantitative investigation on important system measures such as reliability, availability and performance is conducted, and for this purpose a variety of modeling formalisms and solution methods are employed based on the matching of the underlying model assumptions and nature of the system aspects being studied. One of the primary focuses of the model development is on solution efficiency and scalability of the models to large systems. The accuracy of the developed models are validated through extensive simulation.





Xia, Ruofan (2015). Stochastic Modeling of Modern Storage Systems. Dissertation, Duke University. Retrieved from


Dukes student scholarship is made available to the public using a Creative Commons Attribution / Non-commercial / No derivative (CC-BY-NC-ND) license.