CSR: CC: Small: Application-Level Consistency for Cloud-Based Computing
University Of Wisconsin-Madison, Madison WI
Investigators
Abstract
The Wisconsin Crash-Consistency Project (WiCC) attacks the fundamental problem of crash recovery in cloud-based applications. Current web applications are not built in a principled and robust manner, relying on ad hoc implementation techniques and arcane system details to realize correct behavior. Unfortunately, the result of this approach can be chaotic, and, when an unfortunate event such as a system crash or power loss arises, can lead to data loss or corruption. As data lies at the heart of virtually every modern web service, such loss or corruption is a nuisance at best, and catastrophic at worst. The WiCC project addresses these issues in a two-pronged manner. First, WiCC provides tools to analyze current applications (and underlying file systems) and thus improves the current state of the art in crash recovery. These tools exercise the systems-under-test through novel state-space exploration techniques, constructing scenarios where a given system will not function correctly. Such feedback is critical for developers, as they can readily determine problems with the systems they are building. Second, WiCC puts forth a new storage system, StreamFS, that presents new abstractions to ease the construction of correct cloud-based data management applications. Current system interfaces are cumbersome and error-prone; StreamFS reduces the complexity of the file system interface and thus improves the state of the art in robust system construction.
View original record on NSF Award Search →