Home → Magazine Archive → January 2023 (Vol. 66, No. 1) → Research for Practice: Crash Consistency → Abstract

Research for Practice: Crash Consistency

By Ramnatthan Alagappan, Peter Alvaro

Communications of the ACM, Vol. 66 No. 1, Pages 52-54

[article image]

back to top 

For this Research for Practice entry, we asked Ram Alagappan, an assistant professor at the University of Illinois Urbana Champaign, to survey recent research on crash consistency—the guarantee that application data will survive system crashes. Unlike memory consistency, crash consistency is an end-to-end concern, requiring not only that the lower levels of the system (for example, the file system) are implemented correctly, but also that their interfaces are used correctly by applications.

Alagappan has chosen a collection of papers that reflects this complexity, traversing the stack from applications all the way to hardware. The first paper focuses on the file system—upon which applications that hope to provide crash consistency must rely—and uses bug-finding techniques to witness violations of interface-level guarantees. The second moves up the stack, rethinking the interfaces that file systems provide to application programmers to make it easier to write crash-consistent programs. In the last, the plot thickens with the new challenges that persistent memory brings to crash consistency. It explores how to mitigate those challenges using cache-coherent accelerators. I learned a lot reading these selections, and I am sure that you will too.
    —Peter Alvaro


No entries found