先进数据系统实验室

　　我们实验室在基于混合存储介质优化LSM-tree键值存储的工作被存储领域顶级期刊ACM TOS(CCF推荐A类)接收。向各位参与研究工作的老师、同学、合作者表示祝贺。

　　论文题目：Leveraging NVMe SSDs for Building A Fast,Cost-EffectiveLSM-tree Based KV Store

　　论文摘要：Key-Value(KV)stores support many crucial applications and services.They perform fast in-memory process-ing,but are still often limited by I/O performance.The recent emergence of high-speed commodity NVMeSSDs has propelled new KV system designs that take advantage of their ultra-low latency and high bandwidth.Meanwhile,to switch to entirely new data layouts and scale up entire databases to high-end SSDs requiresconsiderable investment.As a compromise,we propose SpanDB,an LSM-tree-based KV store that adapts the popular RocksDB sys-tem to utilizeselective deployment of high-speed SSDs.SpanDB allows users to host the bulk of their data oncheaper and larger SSDs(and even HDDs with certain workloads),while relocating write-ahead logs(WAL)and the top levels of the LSM-tree to a much smaller and faster NVMe SSD.To better utilize this fast disk,SpanDB provides high-speed,parallel WAL writes via SPDK,and enables asynchronous request processingto mitigate inter-thread synchronization overhead and work efficiently with polling-based I/O.To ease thelive data migration between fast and slow disks,we introduce TopFS,a stripped-down file system provid-ing familiar file interface wrappers on top of SPDK I/O.Our evaluation shows that SpanDB simultaneouslyimproves RocksDB’s throughput by up to 8.8×and reduces its latency by 9.5-58.3%.Compared with KVell,a system designed for high-end SSDs,SpanDB achieves 96-140%of its throughput,with a 2.3-21.6×lowerlatency,at a cheaper storage configuration.