{"title":"P2PNIC:网卡间直接通信,实现高速报文转发","authors":"Yukito Ueno, Ryo Nakamura, Yohei Kuga, H. Esaki","doi":"10.1109/INFOCOMWKSHPS51825.2021.9484641","DOIUrl":null,"url":null,"abstract":"Against the background of the contiguous growth of the Internet and data center traffic, the performance requirement for software middleboxes is increasing rapidly. Although their performance has been improved by continuous research and development, their packet forwarding architecture depends on the CPU and the main memory. Thus, their throughput cannot exceed the limit of the memory bandwidth, for instance. To address these limits, we propose a novel packet forwarding architecture called P2PNIC. In P2PNIC, a NIC directly communicates with other NICs through the PCIe interconnect without CPU and main memory involvement, like the inter-linecard communication in a hardware router. To show the feasibility and the performance advantages of P2PNIC architecture, we implemented P2PNIC on a programmable 40 GbE NIC and compared the throughput and latency with TestPMD, which is an application of DPDK. The evaluation shows that P2PNIC achieves 40.37 Mpps for 64-byte packets, which is 1.45 times higher than TestPMD. In addition, P2PNIC shows 36% lower latency than TestPMD for 64-byte packets with 1 Gbps background traffic. The P2PNIC architecture accelerates packet forwarding on a general-purpose server and advances software-based network technologies.","PeriodicalId":109588,"journal":{"name":"IEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-05-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"P2PNIC: High-Speed Packet Forwarding by Direct Communication between NICs\",\"authors\":\"Yukito Ueno, Ryo Nakamura, Yohei Kuga, H. Esaki\",\"doi\":\"10.1109/INFOCOMWKSHPS51825.2021.9484641\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Against the background of the contiguous growth of the Internet and data center traffic, the performance requirement for software middleboxes is increasing rapidly. Although their performance has been improved by continuous research and development, their packet forwarding architecture depends on the CPU and the main memory. Thus, their throughput cannot exceed the limit of the memory bandwidth, for instance. To address these limits, we propose a novel packet forwarding architecture called P2PNIC. In P2PNIC, a NIC directly communicates with other NICs through the PCIe interconnect without CPU and main memory involvement, like the inter-linecard communication in a hardware router. To show the feasibility and the performance advantages of P2PNIC architecture, we implemented P2PNIC on a programmable 40 GbE NIC and compared the throughput and latency with TestPMD, which is an application of DPDK. The evaluation shows that P2PNIC achieves 40.37 Mpps for 64-byte packets, which is 1.45 times higher than TestPMD. In addition, P2PNIC shows 36% lower latency than TestPMD for 64-byte packets with 1 Gbps background traffic. The P2PNIC architecture accelerates packet forwarding on a general-purpose server and advances software-based network technologies.\",\"PeriodicalId\":109588,\"journal\":{\"name\":\"IEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)\",\"volume\":\"51 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-05-10\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/INFOCOMWKSHPS51825.2021.9484641\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/INFOCOMWKSHPS51825.2021.9484641","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
P2PNIC: High-Speed Packet Forwarding by Direct Communication between NICs
Against the background of the contiguous growth of the Internet and data center traffic, the performance requirement for software middleboxes is increasing rapidly. Although their performance has been improved by continuous research and development, their packet forwarding architecture depends on the CPU and the main memory. Thus, their throughput cannot exceed the limit of the memory bandwidth, for instance. To address these limits, we propose a novel packet forwarding architecture called P2PNIC. In P2PNIC, a NIC directly communicates with other NICs through the PCIe interconnect without CPU and main memory involvement, like the inter-linecard communication in a hardware router. To show the feasibility and the performance advantages of P2PNIC architecture, we implemented P2PNIC on a programmable 40 GbE NIC and compared the throughput and latency with TestPMD, which is an application of DPDK. The evaluation shows that P2PNIC achieves 40.37 Mpps for 64-byte packets, which is 1.45 times higher than TestPMD. In addition, P2PNIC shows 36% lower latency than TestPMD for 64-byte packets with 1 Gbps background traffic. The P2PNIC architecture accelerates packet forwarding on a general-purpose server and advances software-based network technologies.