[Fuego] [PATCH 3/3] iperf3: use bps instead of Mbps

[Tim.Bird at sony.com]

> -----Original Message-----
> From: Daniel Sangorrin
> 
> > -----Original Message-----
> > From: fuego-bounces at lists.linuxfoundation.org [mailto:fuego-
> bounces at lists.linuxfoundation.org] On Behalf Of Daniel Sangorrin
> > Sent: Thursday, March 08, 2018 1:46 PM
> > To: Tim.Bird at sony.com; fuego at lists.linuxfoundation.org
> > Subject: Re: [Fuego] [PATCH 3/3] iperf3: use bps instead of Mbps
> >
> > > -----Original Message-----
> > > From: Tim.Bird at sony.com [mailto:Tim.Bird at sony.com]
> > > Sent: Thursday, March 08, 2018 8:37 AM
> > > To: daniel.sangorrin at toshiba.co.jp; fuego at lists.linuxfoundation.org
> > > Subject: RE: [Fuego] [PATCH 3/3] iperf3: use bps instead of Mbps
> > >
> > > A couple of issues I saw while testing this:
> > >
> > > Does the iperf3 server only support one connection at a time?
> > > I tried running multiple tests at the same time, and one of them
> > > got (from the test console log):
> >
> > iperf3 has the -P option for the client to run a test with multiple parallel
> connections.
> > However, it does not support multiple instances of iperf3's client running at
> the same time.
> > # You would need to run it twice on separate ports (-p port)
> > There is a wrapper script called bwctl that ensures that only one test is
> running at a time.
> > I will investigate how to use it and see if that solves the problem.
> 
> I checked bwctl, but it was too much for our purposes. Instread I have added
> retry
> logic to fuego_test.sh for running iperf3 on multiple boards at the same time.
> 
> >
> > > #### doing fuego phase: run ########
> > > Using server ip address: 10.0.1.61
> > > Starting iperf3 server on localhost (SERVER IP: 10.0.1.61)
> > > Starting iperf3 client on the target (CLIENT IP: 10.0.1.91)
> > > {
> > > 	"start":	{
> > > 		"connected":	[],
> > > 		"version":	"iperf 3.1.3",
> > > 		"system_info":	"Linux m3ulcb 4.9.0-yocto-standard #1 SMP
> PREEMPT Wed Jul 5 15:05:42 PDT 2017 aarch64"
> > > 	},
> > > 	"intervals":	[],
> > > 	"end":	{
> > > 	},
> > > 	"error":	"error - unable to connect to server: Connection
> refused"
> > > }
> > >
> > > The second run, without any contention for the server, ran OK.
> > >
> > > Also, most of my board fail the test due to very high CPU utilization.
> > >
> > > What is the expected CPU utilization?
> >
> > It depends on the board I guess.
> >
> > Beaglebone black: 68.3%
> > An intel i7 PC: 10.7%
> > Zynq702: 54.2%
> > Tegrak1 (jetson): 36.8%
> >
> > You can control the amount of bandwidth with the -b option (e.g. -b 100M).
> 
> I have changed the default specs to be a bit more lax about the CPU
> utilization
> and allow 100%.
> 
> >
> > > It would be nice to see the png even in the failure case, as there might
> > > be useful data in it.
> >
> > OK.
> 
> I added the links to the fail case as well.
> 
> >
> > > Overall, patch looks good, but test could use some refinement.
> >
> > Thanks. I just found another problem myself.
> > I was using the stdout JSON file directly, but sometimes other messages
> get
> > intermingled and the parser gives a "can't decode JSON" error. I will fix this
> > and send you another patch.
> 
> This should be fixed in my latest patches. They are on my git repository
> as well for your convenience.
> 
> Daniel

Thanks.  Most of these were applied.
 -- Tim