Re: [Make-wifi-fast] iperf 2.0.14

Re: [Make-wifi-fast] iperf 2.0.14

[Toke Høiland-Jørgensen]

Bob McMahon via Make-wifi-fast <make-wifi-fast@lists.bufferbloat.net>
writes:

> From: Bob McMahon <bob.mcmahon@broadcom.com>
> Subject: iperf 2.0.14
> To: Make-Wifi-fast <make-wifi-fast@lists.bufferbloat.net>
> Date: Fri, 24 Jan 2020 12:52:16 -0800
>
> Hi All,
>
> Just a heads up that iperf 2.0.14 <https://sourceforge.net/projects/iperf2/>
> is getting closer to a release. Feel free to suggest features that you
> might find helpful.

I have an old TODO item sitting around telling me to go over which
features we use from netperf in Flent that we don't currently use for
iperf. So thank you for the reminder; since I'm leaving for holidays
tomorrow, here is a quick overview (i.e., I haven't had time to go over
how many of these are actually missing from iperf, and how many are
there, but we're just not using them):

- We extract the following variables from netperf (-k option):

    output_vars = 'THROUGHPUT,LOCAL_CONG_CONTROL,REMOTE_CONG_CONTROL,' \
                  'TRANSPORT_MSS,LOCAL_TRANSPORT_RETRANS,' \
                  'REMOTE_TRANSPORT_RETRANS,LOCAL_SOCKET_TOS,' \
                  'REMOTE_SOCKET_TOS,DIRECTION,ELAPSED_TIME,PROTOCOL,' \
                  'LOCAL_SEND_SIZE,LOCAL_RECV_SIZE,' \
                  'REMOTE_SEND_SIZE,REMOTE_RECV_SIZE,' \
                  'LOCAL_BYTES_SENT,LOCAL_BYTES_RECVD,' \
                  'REMOTE_BYTES_SENT,REMOTE_BYTES_RECVD'

- We pass in DUMP_TCP_INFO to netperf and capture the output (ref:
  https://github.com/tohojo/flent/issues/13)

- We use the netperf features to set:
  - diffserv markings
  - TCP congestion control
  - socket timeout

- For running the public test servers, the ability to just leave a
  'netserver' instance running indefinitely and let it take care of
  negotiating test setups. I don't think this works that well with
  iperf?



If iperf can replicate all of that, it'll be fairly straight-forward to
make Flent backend-agnostic on the TCP test runner front. How much do
you reckon is currently missing, and what would it take to get that in? :)

-Toke

Re: [Make-wifi-fast] iperf 2.0.14

[Bob McMahon]

[-- Attachment #1: Type: text/plain, Size: 2501 bytes --]

Thanks, let me evaluate this

Bob

On Fri, Jan 24, 2020, 1:25 PM Toke Høiland-Jørgensen <toke@redhat.com>
wrote:

> Bob McMahon via Make-wifi-fast <make-wifi-fast@lists.bufferbloat.net>
> writes:
>
> > From: Bob McMahon <bob.mcmahon@broadcom.com>
> > Subject: iperf 2.0.14
> > To: Make-Wifi-fast <make-wifi-fast@lists.bufferbloat.net>
> > Date: Fri, 24 Jan 2020 12:52:16 -0800
> >
> > Hi All,
> >
> > Just a heads up that iperf 2.0.14 <
> https://sourceforge.net/projects/iperf2/>
> > is getting closer to a release. Feel free to suggest features that you
> > might find helpful.
>
> I have an old TODO item sitting around telling me to go over which
> features we use from netperf in Flent that we don't currently use for
> iperf. So thank you for the reminder; since I'm leaving for holidays
> tomorrow, here is a quick overview (i.e., I haven't had time to go over
> how many of these are actually missing from iperf, and how many are
> there, but we're just not using them):
>
> - We extract the following variables from netperf (-k option):
>
>     output_vars = 'THROUGHPUT,LOCAL_CONG_CONTROL,REMOTE_CONG_CONTROL,' \
>                   'TRANSPORT_MSS,LOCAL_TRANSPORT_RETRANS,' \
>                   'REMOTE_TRANSPORT_RETRANS,LOCAL_SOCKET_TOS,' \
>                   'REMOTE_SOCKET_TOS,DIRECTION,ELAPSED_TIME,PROTOCOL,' \
>                   'LOCAL_SEND_SIZE,LOCAL_RECV_SIZE,' \
>                   'REMOTE_SEND_SIZE,REMOTE_RECV_SIZE,' \
>                   'LOCAL_BYTES_SENT,LOCAL_BYTES_RECVD,' \
>                   'REMOTE_BYTES_SENT,REMOTE_BYTES_RECVD'
>
> - We pass in DUMP_TCP_INFO to netperf and capture the output (ref:
>   https://github.com/tohojo/flent/issues/13)
>
> - We use the netperf features to set:
>   - diffserv markings
>   - TCP congestion control
>   - socket timeout
>
> - For running the public test servers, the ability to just leave a
>   'netserver' instance running indefinitely and let it take care of
>   negotiating test setups. I don't think this works that well with
>   iperf?
>
>
>
> If iperf can replicate all of that, it'll be fairly straight-forward to
> make Flent backend-agnostic on the TCP test runner front. How much do
> you reckon is currently missing, and what would it take to get that in? :)
>
> -Toke
>
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast

[-- Attachment #2: Type: text/html, Size: 3798 bytes --]