Accelerating Distributed Deep Learning using Multi-Path RDMA in Data Center Networks
Accelerating Distributed Deep Learning using Multi-Path RDMA in Data Center Networks
复制标题
DOI：
10.1145/3482898.3483363
复制
发表时间：
2021-10
期刊：
影响因子：
--
通讯作者：
Feng Tian;Yang Zhang;Wei Ye;Cheng Jin;Ziyan Wu;Zhi-Li Zhang
中科院分区：
文献类型：
--
作者：
Feng Tian;Yang Zhang;Wei Ye;Cheng Jin;Ziyan Wu;Zhi-Li Zhang
数据中心网络 (DCN) 广泛部署 RDMA 以支持机器学习等数据密集型应用。虽然 DCN 设计有丰富的多路径拓扑，但当前的 RDMA（硬件）技术不支持多路径传输。在本文中，我们改进了 Maestro（一种纯粹基于软件的多路径 RDMA 解决方案），以有效利用丰富的多路径拓扑来实现负载平衡和可靠性。作为运行在用户空间的“中间件”，Maestro 是模块化和软件定义的：Maestro 将路径选择和负载平衡机制与硬件功能解耦，并允许 DCN 运营商和应用程序根据需要采用最佳机制来做出灵活的决策。因此，可以使用现有 RDMA 硬件 (NIC) 轻松部署 Maestro，以支持分布式深度学习 (DDL) 应用程序。我们的实验表明，Maestro 能够以可忽略的 CPU 开销充分利用多条路径，从而提高 DDL 应用程序的性能。
Data center networks (DCNs) have widely deployed RDMA to support data-intensive applications such as machine learning. While DCNs are designed with rich multi-path topology, current RDMA (hardware) technology does not support multi-path transport. In this paper we advance Maestro- a purely software-basedmulti-path RDMA solution - to effectively utilize the rich multi-path topology for load balancing and reliability. As a "middleware" operating at the user-space, Maestro is modulaR@and software-defined:Maestro decouples path selection and load balancing mechanisms from hardware features, and allows DCN operators and applications to make flexible decisions by employing the best mechanisms as needed. As such, Maestro can be readily deployed using existing RDMA hardware (NICs) to support distributed deep learning (DDL) applications. Our experiments show that Maestro is capable of fully utilizing multiple paths with negligible CPU overheads, thereby enhancing the performance of DDL applications.