Utilizing NVDIMM to alleviate the I/O performance gap for big data workloads

摘要

In this talk, I will present our recent work to utilize NVDIMM (Non-Volatile Dual Inline Memory Module) to bridge the I/O performance gap for big data workloads. Specially, we develop a scheduling scheme in memory controller to minimize the interference between the native and the l/O-derived memory traffic by exploiting both data access criticality and resource utilization. Several mechanisms are designed to better orchestrate traffic between the memory controller and NAND flash to alleviate the speed discrepancy issue. We also propose a proactive GC scheme for the NVDIMM controller and flash controller to intelligently synchronize and transfer data involving in forthcoming GC operations. Our experimental results show that overall the proposed techniques yield 10%∼35% performance improvements over the state-of-the-art baseline schemes.