Characterizing and Optimizing Remote Persistent Memory with RDMA and NVM


Xingda Wei, Xiating Xie, Rong Chen, Haibo Chen, and Binyu Zang, Shanghai Jiao Tong University; Shanghai AI Laboratory; Engineering Research Center for Domain-specific Operating Systems


The appealing properties of NVM including high performance, persistence, and byte-addressability, and a recent active thread of building remote memory systems with RDMA, have produced considerable interest in combining them for fast and persistent remote memory systems. However, many prior systems are either based on emulated NVM or have failed to fully exploit NVM characteristics, leading to suboptimal performance.

This paper conducts a systematic study to summarize optimization hints that the system designer can use to exploit NVM with RDMA better. Specifically, we demonstrate how system configurations, NVM access patterns, and RDMAaware optimizations affect the efficacy of RDMA-NVM systems. Based on the summarized hints, we empirically study the design of two existing RDMA-NVM systems, namely a distributed database (DrTM+H) and a distributed file system (Octopus). Both systems are designed when no production NVM is available, and neither of them achieves good performance on it. Our optimized systems achieve up to 2.4X (from 1.2X) better performance

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

@inproceedings {273889,
author = {Xingda Wei and Xiating Xie and Rong Chen and Haibo Chen and Binyu Zang},
title = {Characterizing and Optimizing Remote Persistent Memory with {RDMA} and {NVM}},
booktitle = {2021 USENIX Annual Technical Conference (USENIX ATC 21)},
year = {2021},
isbn = {978-1-939133-23-6},
pages = {523--536},
url = {},
publisher = {USENIX Association},
month = jul

Presentation Video