Contention and Locality-Aware Work-Stealing for Iterative Applications in Multi-socket Computers.

        Modern large-scale computers have shifted to Multi-socket Multi-core (MSMC) architecture, where multiple CPU chips are integrated into a machine as sockets and multiple memory nodes are integrated into the shared main memory (NUMA). To improve the hardware utilization of MSMC computers, multiple programs are often executed concurrently. However, most existing work-stealing schedulers are designed for single-socket architecture and contention-free scenario. Work-stealing programs suffer from severe remote memory accesses and serious interference from the co-located programs in MSMC architectures, which in turn significantly degrade their performance. To solve the two problems, we propose a Contention and Locality-Aware Work-Stealing (CLAWS) scheduler. CLAWS first evenly distributes the data set of a program to all the memory nodes and allocates a task to the socket where the local memory node stores its data. Then, according to real-time contention situation in each socket collected at runtime, CLAWS dynamically migrates data and re-allocates the corresponding tasks to balance workload and reduce remote memory accesses. Experimental results show that CLAWS can improve the performance of memory-bound programs for 40.1% on average compared with the traditional work-stealing schedulers. Meanwhile, CLAWS is also more energy efficient than traditional work-stealing schedulers.