Method for Enabling RDMA Transport Peer to Peer Transfer with NVMeoF Ethernet SSDs

Venkataratnam Nimmagadda, Sandeep Kumar Ananthapalli
{"title":"Method for Enabling RDMA Transport Peer to Peer Transfer with NVMeoF Ethernet SSDs","authors":"Venkataratnam Nimmagadda, Sandeep Kumar Ananthapalli","doi":"10.1109/VLSIDCS53788.2022.9811431","DOIUrl":null,"url":null,"abstract":"P2P (peer to peer) transfer enables to offload Host CPU & its DRAM, while running critical enterprise applications like replication & snapshot. P2P is inherently possible in the PCIe domain because of the Address based and ID based routing characteristics of the single root PCIe topology. CMB (controller memory buffer) can be utilized in PCIe based NVMe (Non-Volatile Memory Express) SSDs to achieve P2P transfer without involving the Host CPU & DRAM [1]. Now a days Data Centers trending to migrate to a scalable, disaggregated storage transports like NVMeoF (Non-Volatile Memory Express Over Fabrics) [4]. Moreover, peers may reside on different racks of Data Center connected with NVMeoF transport. Currently there is no inherent method available to use CMB feature [5] and doing P2P transfer in NVMeoF. In this paper we approach this problem by using VUC (vendor unique command) for sharing Peer's connection information & establishing a dedicated P2P QP (Queue pair) connection among the peer Ethernet SSDs. Through this method the host CPU will just give control messages to initiate the P2P transfer so that highly intensive IO transfer responsibility is fully offloaded to peer NVMeoF SSDs. The proposed approach is energy efficient and nullifies the power consumed by Host DRAM for data copy & reduces load on ethernet switch to half, while the data copy speed is doubled without affecting the throughput.","PeriodicalId":307414,"journal":{"name":"2022 IEEE VLSI Device Circuit and System (VLSI DCS)","volume":"85 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE VLSI Device Circuit and System (VLSI DCS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/VLSIDCS53788.2022.9811431","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

P2P (peer to peer) transfer enables to offload Host CPU & its DRAM, while running critical enterprise applications like replication & snapshot. P2P is inherently possible in the PCIe domain because of the Address based and ID based routing characteristics of the single root PCIe topology. CMB (controller memory buffer) can be utilized in PCIe based NVMe (Non-Volatile Memory Express) SSDs to achieve P2P transfer without involving the Host CPU & DRAM [1]. Now a days Data Centers trending to migrate to a scalable, disaggregated storage transports like NVMeoF (Non-Volatile Memory Express Over Fabrics) [4]. Moreover, peers may reside on different racks of Data Center connected with NVMeoF transport. Currently there is no inherent method available to use CMB feature [5] and doing P2P transfer in NVMeoF. In this paper we approach this problem by using VUC (vendor unique command) for sharing Peer's connection information & establishing a dedicated P2P QP (Queue pair) connection among the peer Ethernet SSDs. Through this method the host CPU will just give control messages to initiate the P2P transfer so that highly intensive IO transfer responsibility is fully offloaded to peer NVMeoF SSDs. The proposed approach is energy efficient and nullifies the power consumed by Host DRAM for data copy & reduces load on ethernet switch to half, while the data copy speed is doubled without affecting the throughput.
使用NVMeoF以太网ssd开启RDMA传输对等传输的方法
P2P(点对点)传输可以卸载主机CPU及其DRAM,同时运行关键的企业应用程序,如复制和快照。由于单根PCIe拓扑的基于地址和基于ID的路由特性,P2P在PCIe域中是可能的。CMB(控制器内存缓冲区)可以在基于PCIe的NVMe (Non-Volatile memory Express) ssd中使用,实现P2P传输,而无需涉及主机CPU和DRAM[1]。现在,数据中心倾向于迁移到可扩展的、可分解的存储传输,如NVMeoF(非易失性存储器快速传输)[4]。此外,对等体可能位于通过NVMeoF传输连接的数据中心的不同机架上。在NVMeoF中利用CMB特性[5]进行P2P传输,目前还没有固有的方法可用。本文通过使用VUC(供应商唯一命令)来共享对等体的连接信息,并在对等体以太网ssd之间建立专用的P2P QP(队列对)连接来解决这个问题。通过这种方法,主机CPU只会给出控制消息来启动P2P传输,从而将高度密集的IO传输责任完全卸载给对等的NVMeoF ssd。所提出的方法节能,消除了主机DRAM用于数据复制的功耗,并将以太网交换机的负载减少了一半,同时数据复制速度提高了一倍,而不影响吞吐量。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信