NyxCache: Flexible and Efficient Multi-tenant Persistent Memory Caching

Authors: 

Kan Wu, Kaiwei Tu, and Yuvraj Patel, University of Wisconsin–Madison; Rathijit Sen and Kwanghyun Park, Microsoft; Andrea Arpaci-Dusseau and Remzi Arpaci-Dusseau, University of Wisconsin–Madison

Abstract: 

We present NyxCache (Nyx), an access regulation framework for multi-tenant persistent memory (PM) caching that supports light-weight access regulation, per-cache resource usage estimation and inter-cache interference analysis. With these mechanisms and existing admission control and capacity allocation logic, we build important sharing policies such as resource-limiting, QoS-awareness, fair slowdown, and proportional sharing: Nyx resource-limiting can accurately limit PM usage of each cache, providing up to 5x better performance isolation than a bandwidth-limiting method. Nyx QoS can provide QoS guarantees to latency-critical caches while providing higher throughput (up to 6x vs. previous DRAM-based approaches) to best-effort caches that are not interfering. Finally, we show that Nyx is useful for realistic workloads, isolating write-spikes and ensuring that important caches are not slowed down by increased best-effort traffic.

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.

This content is available to:

BibTeX
@inproceedings {277850,
title = {{NyxCache}: Flexible and Efficient Multi-tenant Persistent Memory Caching},
booktitle = {20th USENIX Conference on File and Storage Technologies (FAST 22)},
year = {2022},
isbn = {978-1-939133-26-7},
address = {Santa Clara, CA},
pages = {1--16},
url = {https://www.usenix.org/conference/fast22/presentation/wu},
publisher = {USENIX Association},
month = feb,
}

Presentation Video