HatRPC: Hint-Accelerated Thrift RPC over RDMA

Authors: Tianxi Li and Haiyang Shi (Ohio State University) and Xiaoyi Lu (University of California, Merced)

Abstract: In this paper, we propose a novel hint-accelerated RPC framework based on Apache Thrift over Remote Direct Memory Access (RDMA) protocols, called HatRPC. HatRPC adopts a hierarchical hint design towards optimizing heterogeneous RPC services and functions. The proposed hint design is composed of service-granularity and function-granularity hints for achieving varied optimization goals and reduced design space for further optimizing the underneath RDMA communication engine. The effectiveness and efficiency of HatRPC are validated and evaluated with our proposed Apache Thrift Benchmarks (ATB), and an RDMA-enhanced commercial database system. We also co-design a key-value store called HatKV with LMDB and evaluate it with YCSB workload. Performance evaluations show that the proposed HatRPC approach can deliver up to 55% performance improvement for ATB benchmarks and up to 1.51x speedup for TPC-H queries compared with vanilla Thrift over IPoIB. In addition, the co-designed HatKV can achieve up to 85.5% enhancement for YCSB workload.


Presentation: file


Back to Technical Papers Archive Listing