Three trends will dominate the storage systems of tomorrow: increasingly massive amounts of data, the incredible growth of software complexity, and the increasing use of cheap and less reliable hardware. These trends present us with a huge challenge: How can we promise users that storage systems work robustly in spite of their massive software complexity and the broad range of hardware failures that can arise? Addressing this question is not straightforward as current approaches scatter recovery code in thousands of lines of intricate, low-level C code. As a result, reliability problems are often found in current storage systems.

In this talk, I will present how we build a new generation of more robust and reliable storage systems via simpler designs and powerful testing frameworks. Specifically, I will first describe I/O Shepherding and SQCK, new online and offline reliability frameworks for OS-level file systems, with which we advocate a higher-level strategy where the logic of reliability policies can be described clearly and concisely. I will then present my recent work in improving cloud storage reliability with FATE and DESTINI, a powerful failure testing service and a framework for declarative recovery specifications. Finally, I will close this talk with my future plans in the area of cloud storage and new storage technology.