paperbot · PL 论文追踪

全部论文 待读 收藏

P2 The benefits of general-purpose on-NIC memory

通用 on-NIC 内存的优势
ASPLOS Proceedings of the 27th ACM International Conference on Architectural Support for Programming Languages and Operating Systems2022匹配度 0/10 · 该论文研究网卡内存(on-NIC memory)在 NFV 和键值存储中的系统与体系结构优化,属于网络/系统领域,与用户关注的静态分析、LLM/Agent、浮点程序分析、贝叶斯程序分析等方向没有直接交集。引用 44 inbox
Boris Pismenny, Liran Liss, Adam Morrison, Dan Tsafrir
crossref: 2026-09-12dblp: 2026-09-12

AI 速览

一句话:提出通用 on-NIC 内存方案,通过将数据与热项驻留网卡内存,显著加速 NFV 与键值存储。

随着网卡性能快速提升,新出现的片上/板载网卡内存(nicmem)未被通用软件有效利用。本文提出将 nicmem 作为通用内存,并针对两类负载设计机制:对 NFV,将包头与数据分离,尽量把数据留在 nicmem,减少 PCIe 流量、内存带宽和 CPU 处理;对键值存储,利用访问偏斜,让网络栈把频繁访问项保留在 nicmem。评估显示,NFV 延迟最多降低 23%、吞吐提升 19%;memcached 延迟最多降低 43%、吞吐提升 80%。

问题
网卡性能快速增长,但其新增的通用片上/板载内存未被利用,NFV 与键值存储又受 PCIe 流量、内存带宽和 CPU 开销制约。
方法
提出通用 nicmem 方案:NFV 中分离包头与数据并尽量将数据留在 nicmem;键值存储中利用访问偏斜,通过网络栈机制让热项驻留 nicmem。
评估
在 NFV 与 memcached 负载上测量延迟和吞吐,最高分别取得 23%/19% 与 43%/80% 的延迟降低/吞吐提升。
原文摘要(Abstract)

We propose to use the small, newly available on-NIC memory ("nicmem") to keep pace with the rapidly increasing performance of NICs. We motivate our proposal by accelerating two types of workload classes: NFV and key-value stores. As NFV workloads frequently operate on headers---rather than data---of incoming packets, we introduce a new packet-processing architecture that splits between the two, keeping the data on nicmem when possible and thus reducing PCIe traffic, memory bandwidth, and CPU processing time. Our approach consequently shortens NFV latency by up to 23% and increases its throughput by up to 19%. Similarly, because key-value stores commonly exhibit skewed distributions, we introduce a new network stack mechanism that lets applications keep frequently accessed items on nicmem. Our design shortens memcached latency by up to 43% and increases its throughput by up to 80%.

链接与引用

DOI 原文 · DBLP

BibTeX
@inproceedings{PismennyL0T22,
  title = {The benefits of general-purpose on-NIC memory},
  author = {Boris Pismenny and Liran Liss and Adam Morrison and Dan Tsafrir},
  booktitle = {ASPLOS},
  year = {2022},
  doi = {10.1145/3503222.3507711}
}