Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

YouTube

KRR - Efficient and Scalable Kernel Record Replay

USENIX via YouTube

Overview

Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about KRR, a novel kernel record-replay framework designed to efficiently capture and reproduce kernel failures in this 16-minute conference presentation from OSDI '25. Discover how researchers from SmartX and Purdue University developed a solution to address the challenges of diagnosing complex kernel bugs that are difficult to reproduce in deployment environments. Explore the innovative split-recorder design that employs both guest and host systems to jointly record kernel execution, significantly narrowing the scope compared to traditional whole-stack approaches. Examine performance benchmarks showing KRR's superior scalability up to 8 cores with only 1.52× to 2.79× slowdown compared to native execution, while traditional whole-VM record-replay systems suffer from 8.97× to 29.94× slowdown across workloads including kernel compilation, RocksDB, and Nginx. Review the practical validation results demonstrating KRR's effectiveness in reproducing 17 real-world Linux bugs across different versions, including 6 non-deterministic bugs and 5 high-risk CVEs, with successful recording and reproduction of all but one non-deterministic bug.

Syllabus

OSDI '25 - KRR: Efficient and Scalable Kernel Record Replay

Taught by

USENIX

Reviews

Start your review of KRR - Efficient and Scalable Kernel Record Replay

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.