Next Generation Reader–Writer Lock Design: Needs, Challenges and Innovation

Reader–writer (RW) locks allow concurrent access for read-only operations and enforce exclusive access for write operations.  However, traditional RW locks suffer from inefficiencies that become particularly pronounced on highly concurrent, resource-constrained mobile devices: Spin-waiting under contention and imprecise wake-ups waste cycles and energy, writer starvation harms real-time operations, and per-lock metadata strains limited memory resources.

To overcome these limitations, in this talk we propose a compact, fair, and efficient RW lock design that addresses these shortcomings by improving scalability, fairness, and real-time responsiveness for mobile workloads. Concretely, we compress lock state to minimize the memory footprint, eliminate spinning entirely during contention, and enforce strict writer priority to prevent starvation.  As a basis for our design, we leverage an existing futex-based MCS queue.