Skip to Content.
Sympa Menu

perfsonar-user - [perfsonar-user] Re: Iperf3 + bandwidth for TCP

Subject: perfSONAR User Q&A and Other Discussion

List archive

[perfsonar-user] Re: Iperf3 + bandwidth for TCP


Chronological Thread 
  • From: Aaron Wood <>
  • To:
  • Cc: Arman Mohsin <>, , "" <>, "" <>
  • Subject: [perfsonar-user] Re: Iperf3 + bandwidth for TCP
  • Date: Sun, 14 May 2017 16:53:20 -0700
  • Ironport-phdr: 9a23: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

Iperf3 (in the stable branches), performs user-level pacing by using a 100ms timer.  Master uses a 1ms timer (and can be set down to 1µs with the --pacing-timer option).  My experience is that the 1ms timer greatly improves smoothness (and can lower drop rates on UDP, especially if you're restricting to a known bottleneck that isn't the local interface).

But user-level pacing works at the "good-put" level, not at the wire-rate.  Whereas when fq-pacing is in use, that's in terms of MAC frames, so as much as 60 bytes larger per frame.

-Aaron

On Sun, May 14, 2017 at 12:18 PM, Brian Tierney <> wrote:

-b works for both TCP and UDP, but --fq-rate works much better for TCP on CentOS7/Debian8 based hosts, as it does kernel-level pacing, not user-level pacing.



On Sat, May 13, 2017 at 7:54 PM, Arman Mohsin <> wrote:
Hi Brian,

Hope all is well. Does the iperf3 -c -b option limit bandwidth for TCP as well or is it ONLY for UDP? I tried to read the documentation in the man page but it is a bit confusing. In the earlier versions of iperf the -b option is used to test only UDP but I am not sure if it supports in the latest version (iperf3)


Regards,

Arman



--
Brian Tierney, http://www.es.net/tierney
Energy Sciences Network (ESnet), Berkeley National Lab
http://fasterdata.es.net

--
You received this message because you are subscribed to the Google Groups "iperf-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to .
To post to this group, send email to .
Visit this group at https://groups.google.com/group/iperf-dev.
For more options, visit https://groups.google.com/d/optout.




Archive powered by MHonArc 2.6.19.

Top of Page