From: "Xing, Beilei" <beilei.xing@intel.com>
To: jiangheng <jiangheng14@huawei.com>
Cc: "users@dpdk.org" <users@dpdk.org>,
"Zhang, Yuying" <yuying.zhang@intel.com>
Subject: RE: the packet delay is too long when I used dpdk i40e NIC sent single packet
Date: Wed, 12 Jul 2023 06:21:31 +0000 [thread overview]
Message-ID: <LV2PR11MB59973DC8E56601D9E7F81027F736A@LV2PR11MB5997.namprd11.prod.outlook.com> (raw)
In-Reply-To: <20230711072800.223661-1-jiangheng14@huawei.com>
Please refer to the INTERNAL description in 710 datasheet.
Interval
ITR 'n' interval, where 'n' is the register index = 0,1,2 for the three ITRs per interrupt.
It is defined in 2 us units enabling interval range from zero to 8160 us (0xFF0).
Setting the INTERVAL to zero enables immediate interrupt.
This register can also be programmed also by setting the INTERVAL field in the
matched xxINT_DYN_CTLx register.
BR,
Beilei
> -----Original Message-----
> From: jiangheng <jiangheng14@huawei.com>
> Sent: Tuesday, July 11, 2023 3:28 PM
> To: Xing, Beilei <beilei.xing@intel.com>
> Cc: users@dpdk.org; jiangheng14@huawei.com; Zhang, Yuying
> <yuying.zhang@intel.com>
> Subject: the packet delay is too long when I used dpdk i40e NIC sent single
> packet
>
> Hi,
> I changed I40E_QUEUE_ITR_INTERVAL_DEFAULT to 10us, latency is
> significantly lower.
> what is the effect of modifying this parameter?
>
> > Hi,
> >
> > Could you try to change interval configuration to check if it helps?
> > Please refer to i40e_calc_itr_interval function.
> > In theory, the lower interval, the lower latency.
> >
> > BR,
> > Beilei
> >
> >> -----Original Message-----
> >> From: jiangheng <jiangheng14@huawei.com>
> >> Sent: Wednesday, July 5, 2023 9:33 AM
> >> To: Zhang, Yuying <yuying.zhang@intel.com>; Xing, Beilei
> >> <beilei.xing@intel.com>
> >> Cc: users@dpdk.org
> >> Subject: the packet delay is too long when I used dpdk i40e NIC sent
> >> single packet
> >>
> >> Hi Beilei:
> >> we used dpdk implement a user mode tcp protocol stack, and we are
> >> testing the delay of a single packet.
> >> however, on i40e NIC(XL710), the packet delay is higher than that of the
> kernel.
> >> After some tests, we find that the time consuming may be the pahse
> >> when i40e sends packet.(period from rte_eth_tx_burst to packet be
> >> sent) The same test is performed on the ixgbe(82599) NIC. the delay is
> lower than that in kernel.
> >>
> >> I used dpdk-21.11 stable version.
> >> I tried upgrading the i40e's firmware version to 9.10, but there was
> >> no improvement.
> >> Any suggestion?
> >> Thank you.
> >>
> >>
> >>
> >> The following is packet delay statistics of tcpdump:
> >> i40e NIC:
> >> 410 10.373817 192.168.4.4 192.168.4.5 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=9642 Win=1048384 Len=0
> >> 10us
> >> 411 10.373827 192.168.4.4 192.168.4.5 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=9642 Ack=6511 Win=1048512 Len=36
> >> 27us
> >> 412 10.373854 192.168.4.5 192.168.4.4 TCP 60 8877
> >> →58312 [ACK] Seq=6511 Ack=9678 Win=1048320 Len=0
> >> 9us
> >> 413 10.373863 192.168.4.5 192.168.4.4 TCP 90
> 58312
> >> →8877 [PSH, ACK] Seq=9642 Ack=6511 Win=1048512 Len=36
> >> 28us
> >> 414 10.373891 192.168.4.4 192.168.4.5 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=9678 Win=1048320 Len=0
> >> 11us
> >> 415 10.373902 192.168.4.4 192.168.4.5 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=9678 Ack=6511 Win=1048512 Len=36
> >> 28us
> >> 416 10.373930 192.168.4.5 192.168.4.4 TCP 60 8877
> >> →58312 [ACK] Seq=6511 Ack=9714 Win=1048320 Len=0
> >> 11us
> >> 417 10.373941 192.168.4.5 192.168.4.4 TCP 90
> 58312
> >> →8877 [PSH, ACK] Seq=9678 Ack=6511 Win=1048512 Len=36
> >> 25us
> >> 418 10.373967 192.168.4.4 192.168.4.5 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=9714 Win=1048320 Len=0
> >> 11us
> >> 419 10.373978 192.168.4.4 192.168.4.5 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=9714 Ack=6511 Win=1048512 Len=36
> >> 4us
> >> 420 10.373982 192.168.4.5 192.168.4.4 TCP 60 8877
> >> →58312 [ACK] Seq=6511 Ack=9750 Win=1048256 Len=0
> >>
> >>
> >> ixgbe NIC:
> >> 110 13.235208 192.168.4.12 192.168.4.11 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=7266 Win=1047808 Len=0
> >> 10us
> >> 111 13.235218 192.168.4.12 192.168.4.11 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=7266 Ack=6511 Win=1048512 Len=36
> >> 6us
> >> 112 13.235224 192.168.4.11 192.168.4.12 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=7302 Win=1047744 Len=0
> >> 10us
> >> 113 13.235234 192.168.4.11 192.168.4.12 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=7266 Ack=6511 Win=1048512 Len=36
> >> 6us
> >> 114 13.235240 192.168.4.12 192.168.4.11 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=7302 Win=1047744 Len=0
> >> 11us
> >> 115 13.235251 192.168.4.12 192.168.4.11 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=7302 Ack=6511 Win=1048512 Len=36
> >> 5us
> >> 116 13.235256 192.168.4.11 192.168.4.12 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=7338 Win=1047744 Len=0
> >> 11us
> >> 117 13.235267 192.168.4.11 192.168.4.12 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=7302 Ack=6511 Win=1048512 Len=36
> >> 6us
> >> 118 13.235273 192.168.4.12 192.168.4.11 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=7338 Win=1047744 Len=0
> >> 10us
> >> 119 13.235283 192.168.4.12 192.168.4.11 TCP 94
> 58312
> >> →8877 [PSH, ACK] Seq=7338 Ack=6511 Win=1048512 Len=36
> >> 6us
> >> 120 13.235289 192.168.4.11 192.168.4.12 TCP 64 8877
> >> →58312 [ACK] Seq=6511 Ack=7374 Win=1047680 Len=0
next prev parent reply other threads:[~2023-07-12 6:21 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-07-05 1:32 jiangheng
2023-07-06 3:10 ` Xing, Beilei
2023-07-11 7:28 ` jiangheng
2023-07-12 6:21 ` Xing, Beilei [this message]
-- strict thread matches above, loose matches on Subject: below --
2023-07-03 4:00 jiangheng
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=LV2PR11MB59973DC8E56601D9E7F81027F736A@LV2PR11MB5997.namprd11.prod.outlook.com \
--to=beilei.xing@intel.com \
--cc=jiangheng14@huawei.com \
--cc=users@dpdk.org \
--cc=yuying.zhang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).