A Flexible and Scalable Affinity Lock for the Kernel

A number of NUMA-aware synchronization algorithms have been proposed lately to stress the scalability inefficiencies of existing locks. However their presupposed local lock granularity, a physical processor, is often not the optimum configuration for various workloads. This paper further explores the design space by taking into consideration the physical affinity between the cores within a single processor, and presents FSL to support variable and finely tuned group size for different lock contexts and instances. The new design provides a uniform model for the discussion of affinity locks and can completely subsume the previous NUMA-aware designs because they have only discussed one special case of the model. The interfaces of the new scheme are kernel-compatible and thus largely facilitate kernel incorporation. The investigation with the lock shows that an affinity lock with optimal local lock granularity can outperform its NUMA-aware counterpart by 29.40% and 58.28% at 80 cores with different workloads.