Lockify: Understanding Linux Distributed Lock Management Overheads in Shared Storage
summary
Lockify implements a kernel DLM change in Linux 6.6.23 that replaces owner queries during file/directory creation with self-owner notifications and asynchronous ownership management, plus an extended acquisition flag and a write list for retransmit/ack. Evaluated on GFS2 and OCFS2 over MVM/TCP with MDTest, Postmark and Filebench, it shows up to 6.4× throughput improvement in low-contention creation bursts, up to 5.4× under high contention, ~2.0× on Postmark and ~2.5× on Filebench, with crash tests preserving consistency.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:40Shared storage and DLMs
- 1:47How distributed lock management works
- 3:56Motivation: Low contention performance
- 5:01GFS2 creation bottlenecks
- 6:07Remote directory node access root cause
- 7:35Lockify solution
- 9:47Implementation and evaluation
- 10:53Performance results comparison
- 12:44Summary
related talks
40mKernel Locking Engineering
20mWas file locking ever meant to work?
31mEnhancing Scalability of the Vmalloc Mechanism in the Linux Kernel
40mUpdate on Landlock: Lifting the File Reparenting Limits and Supporting Network Rules
20mDLOS: Effective Static Detection of Deadlocks in OS Kernels
01h 09mNew locks for the old kernel
44mShip your Critical Section, Not Your Data: Enabling Transparent Delegation with...
23mProtecting your resources with sanlock