An Analysis of Bugs In Persistent Memory Application

TL;DR

This paper evaluates an open-source bug detection tool for persistent memory applications, discovers new bugs in the PMDK library, and proposes a Deep-Q Learning heuristic to enhance bug search efficiency.

Contribution

It introduces a validation tool that finds new bugs in NVM applications and improves search strategies with a Deep-Q Learning heuristic.

Findings

Discovered 65 new NVM bugs in PMDK library.

Outperformed WITCHER framework in bug detection.

Proposed a Deep-Q Learning heuristic for efficient bug search.

Abstract

Over the years of challenges on detecting the crash consistency of non-volatile persistent memory (PM) bugs and developing new tools to identify those bugs are quite stretching due to its inconsistent behavior on the file or storage systems. In this paper, we evaluated an open-sourced automatic bug detector tool (i.e. AGAMOTTO) to test NVM level hashing PM application to identify performance and correctness PM bugs in the persistent (main) memory. Furthermore, our faithful validation tool able to discovered 65 new NVM level hashing bugs on PMDK library and it outperformed the number of bugs (i.e. 40 bugs) that WITCHER framework was able to identified. Finally, we will propose a Deep-Q Learning search heuristic algorithm over the PM-Aware search algorithm in the state selection process to improve the searching strategy efficiently.