Low speed

[LinSir1997]

The bridging tap network cards, with a maximum transmission rate of only 1.5Gbps, cannot make full use of multi-core cpu,tap network cards support multiple queues(IFF_MULTI_QUEUE),You can use multi-threaded queue to send and receive, improve the performance of the tap network card.
The performance of the single-core cpu cannot meet the requirements of enterprise-level networking, and the fullmesh mode is not supported.

[chipitsine]

Please provide detailed description or repro steps

…

On Tue, Jan 16, 2024, 05:50 LinSir1997 ***@***.***> wrote: The bridging tap network cards, with a maximum transmission rate of only 1.5Gbps, cannot make full use of multi-core cpu,tap network cards support multiple queues(IFF_MULTI_QUEUE),You can use multi-threaded queue to send and receive, improve the performance of the tap network card. The performance of the single-core cpu cannot meet the requirements of enterprise-level networking, and the fullmesh mode is not supported. — Reply to this email directly, view it on GitHub <#1944>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAQ5KUDGGOWW7X42BDBCJR3YOYBKHAVCNFSM6AAAAABB4FWBOGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGA4DGMBRGQ4TANA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[LinSir1997]

Please provide detailed description or repro steps
…
On Tue, Jan 16, 2024, 05:50 LinSir1997 @.> wrote: The bridging tap network cards, with a maximum transmission rate of only 1.5Gbps, cannot make full use of multi-core cpu,tap network cards support multiple queues(IFF_MULTI_QUEUE),You can use multi-threaded queue to send and receive, improve the performance of the tap network card. The performance of the single-core cpu cannot meet the requirements of enterprise-level networking, and the fullmesh mode is not supported. — Reply to this email directly, view it on GitHub <#1944>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQ5KUDGGOWW7X42BDBCJR3YOYBKHAVCNFSM6AAAAABB4FWBOGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGA4DGMBRGQ4TANA . You are receiving this because you are subscribed to this thread.Message ID: @.>

This is a case of velocity measurement in a cascade tunnel using iperf3

The 8-core cpu is not fully utilized, only brbridge threads are sending and receiving messages, and the multi-queue function of tap network card is not brought into play. multi-core can be bound through multi-queue to achieve maximum message sending and receiving performance.
My optimization idea is to use IFF_MULTI_QUEUE to support multi-queue tap network cards, set up multiple fd bound network cards, and set up the sending and receiving of network cards in each thread. in fact, I did the same, but the problem is that I use multiple threads, and the session of each thread is the same as the session of brbridge thread. I encountered the problem of multi-thread crash and there are too many places to modify.