* [dpdk-dev] DPDKthe packet capture performance increases steeply between 11 Rx queues and 12 Rx queues. Any idea why?
[not found] <CAAh6R8Jc9Wp6EfGTR36s4ZT4xeVriGeUE+=BhswaOAtu2ztkpw@mail.gmail.com>
@ 2021-03-24 9:04 ` Rajesh R
0 siblings, 0 replies; only message in thread
From: Rajesh R @ 2021-03-24 9:04 UTC (permalink / raw)
To: dev
Hi,
We are experimenting with testpmd packet generation and capture at 100 Gbps
on two machines back to back. We are observing a peculiar behaviour
that the packet capture performance increases steeply between 11 Rx queues
and 12 Rx queues. Any idea why?
The details of the setup are given below:
DPDK version: *19.11*
NIC: *Mellanox ConnectX-5 100 GbE *
OS: *RHEL 8.2(kernel *4.18.0-193)
Application: *Testpmd(txonly on generation side and rxonly on capture side)*
*It is observed that *capture rate increases almost gradually upto* 11 Rx
queues and then increases steeply *to reach line rate *at 12 queues. The
maximum capture rate with 11 Rx queues is around 80 Gbps and with 12 Rx
queues is 99. Gbps.*
*Is it something to do with DPDK ? Or Mellanox specific?*
--
Regards
Rajesh R
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2021-03-25 13:56 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <CAAh6R8Jc9Wp6EfGTR36s4ZT4xeVriGeUE+=BhswaOAtu2ztkpw@mail.gmail.com>
2021-03-24 9:04 ` [dpdk-dev] DPDKthe packet capture performance increases steeply between 11 Rx queues and 12 Rx queues. Any idea why? Rajesh R
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).