Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about PageFlex, a novel framework for delegating Linux paging policies to user space through this 12-minute conference presentation from USENIX ATC '25. Discover how researchers from Google, UC San Diego, and University of Washington developed a solution to address the challenges hyperscalers face when implementing memory efficiency improvements through transparent demotion of "cold" application data to cheaper memory tiers like compressed memory and NVMe SSDs. Explore how PageFlex leverages eBPF to delegate policy decisions while maintaining low-overhead access to in-kernel memory state and access information, achieving the optimal balance between flexibility and performance. Understand the framework's capability to support different paging strategies for distinct memory regions and application phases, enabling compatibility with existing real-world deployments without requiring application code modifications. Examine the performance results demonstrating how PageFlex can delegate existing kernel-based policies with minimal application slowdown (less than 1%) while effectively realizing benefits from state-of-the-art policies like Hyperbolic caching and Leap prefetching, and unlocking application-specific advantages through region- and phase-aware policy specialization.
