P2PNIC: High-Speed Packet Forwarding by Direct Communication between NICs

Ueno, Yukito; Nakamura, Ryô; Kuga, Yohei; Esaki, Hiroshi

doi:10.1109/infocomwkshps51825.2021.9484641

Cited by 3 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, we have implemented some variants of P2PNIC architecture on Netronome's SmartNICs [34] and conducted an evaluation by comparing them with the DPDK applications as examples of CPU-driven architecture. We extend our previous work [43], which proposed the basic concept of the P2PNIC architecture; whereas the previous method supports only forwarding packets to a single fixed port, this extended version supports each NIC dynamically determining the destination port to which to forward packets from multiple ports by IP routing. Moreover, we have added measurements for the extension with several traffic patterns that load multiple ports simultaneously and confirmed that its throughput scales up to 160 Gbps at maximum.…”

Section: Introductionmentioning

confidence: 99%

A NIC-driven Architecture for High-speed IP Packet Forwarding on General-purpose Servers

Ueno

Nakamura

Kuga

et al. 2022

Journal of Information Processing

Self Cite

View full text Add to dashboard Cite

We propose a high-speed packet forwarding architecture on general-purpose servers, in which a Network Interface Card (NIC) drives packet forwarding by direct packet transfer to other NICs via a PCIe switch. The demand for high-speed packet forwarding technology on general-purpose servers is increasing with the spread of networking concepts such as Network Function Virtualization (NFV). However, the current architecture, which processes packets by CPU, cannot achieve the similar degree of performance that hardware routers can provide because the processing capacity of the CPU and the bandwidth of the main memory constrain the performance. Our proposed method, called P2PNIC, overcomes this constraint by eliminating the CPU and the main memory from the entire packet forwarding. In the P2PNIC architecture, a NIC determines to which NIC to forward the packets and directly transfers the packets to the NIC over the PCIe. We evaluate the P2PNIC architecture by comparing it with the DPDK applications as examples of the current architecture. The evaluation shows that the P2PNIC architecture achieves 3.44 times higher throughput and up to 79% lower latency than the DPDK applications. This study offers a new approach in software-based network infrastructure for achieving comparable performance with hardware routers in the future.

show abstract

Section: Introductionmentioning

confidence: 99%

A NIC-driven Architecture for High-speed IP Packet Forwarding on General-purpose Servers

Ueno

Nakamura

Kuga

et al. 2022

Journal of Information Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…(2) The scheme makes full use of DPDK's parallel packet processing support for multicore and multi-NICs. It adopts a Q-learning data stream scheduling that not only enables the bandwidth overlap of multiple NICs but also adapts to the high speed, high concurrency, high traffic and strong real-time characteristics of communication systems to obtain a higher throughput at a lower cost [5][6][7][8].…”

Section: Introductionmentioning

confidence: 99%

An Optimization Method of Large-Scale Video Stream Concurrent Transmission for Edge Computing

2023

View full text Add to dashboard Cite

Concurrent access to large-scale video data streams in edge computing is an important application scenario that currently faces a high cost of network access equipment and high data packet loss rate. To solve this problem, a low-cost link aggregation video stream data concurrent transmission method is proposed. Data Plane Development Kit (DPDK) technology supports the concurrent receiving and forwarding function of multiple Network Interface Cards (NICs). The Q-learning data stream scheduling model is proposed to solve the load scheduling of multiple queues of multiple NICs. The Central Processing Unit (CPU) transmission processing unit was dynamically selected by data stream classification, as well as a reward function, to achieve the dynamic load balancing of data stream transmission. The experiments conducted demonstrate that this method expands the bandwidth by 3.6 times over the benchmark scheme for a single network port, and reduces the average CPU load ratio by 18%. Compared to the UDP and DPDK schemes, it lowers the average system latency by 21%, reduces the data transmission packet loss rate by 0.48%, and improves the overall system transmission throughput. This transmission optimization scheme can be applied in data centers and edge computing clusters to improve the communication performance of big data processing.

show abstract