2014 Second International Symposium on Computing and Networking (CANDAR) (2014)
Dec. 10, 2014 to Dec. 12, 2014
The operating system storage stack is an important software component, but it faces several reliability threats. The research community has come up with many solutions to address individual parts of this reliability problem. However, when considering the bigger picture of constructing a highly reliable storage stack out of these individual solutions, new questions arise regarding the feasibility, complexity, reliability, and performance of such a combination. In previous works, we have designed a new storage stack called Loris, and developed several individual reliability improvements for it. In this work, we investigate the integration of these improvements with each other and with new functionality, in two steps. First, we add new virtualization extensions to Loris which challenge assumptions we made in our previous work on reliability. Second, we combine all our extensions to form a reliable, virtualizing storage stack. We evaluate the resulting stack in terms of performance and reliability.
Virtualization, Computer crashes, Software reliability, Physical layer, Operating systems
D. C. Moolenbroek, R. Appuswamy and A. S. Tanenbaum, "Putting the Pieces Together: The Construction of a Reliable Virtualizing Object-Based Storage Stack," 2014 Second International Symposium on Computing and Networking (CANDAR), Shizuoka, Japan, 2014, pp. 76-85.