[ros-users] Is ROS limiting channel bandwidth over network?

Eric Perko wisesage5001 at gmail.com
Thu Mar 10 19:44:47 UTC 2011


Sean,

On Thu, Mar 10, 2011 at 2:12 PM, Sean Anderson <
sean.william.anderson at gmail.com> wrote:

> Hi Blaise,
>
> 4MBps is sufficient for sending the compressed stream, the problem is that
> when we were subscribing to only a few compressed image channels through our
> ROS nodes we found that our bandwidth flat lined at 1MBps.
>
> When we subscribe to the raw image stream we see up to 4MBps, this
> initially suggested to me that the compression was too expensive and could
> only manage 2 images being compressed at 5Hz, except that our laptop (dual
> core 2GHz) was not CPU bottlenecked. So there is a bottleneck occurring
> somewhere else in the system. CPU is running at 60% and Wireless N is
> running at 1MBps.
>

Just wondering, but what bandwidth do you see over a LAN connection running
the exact same commands? Might help narrow down where the bottleneck
actually is.

Also, just to rule out the wifi as the culprit, have you tried running the
iperf ( http://iperf.sourceforge.net/ ) tool to get some performance numbers
for your wifi? On one computer run `iperf -s` and on another run `iperf -c
<others_ip_address>`. After a little bit it should output a bandwidth number
on each computer. How does that number compare to the bandwidth you see when
looking at, e.g., rostopic bw for the compressed topics?

- Eric


>
> Thanks so much for you time,
>
> Sean Anderson
>
>
> On Thu, Mar 10, 2011 at 1:03 PM, Blaise Gassend <blaise at willowgarage.com>wrote:
>
>> Hi Sean,
>>
>> Where are you running "rostopic bw"? If you are running it off-robot,
>> then each additional "rostopic bw" will cause the same information to
>> be transferred one extra time over the network. Hence, if you are
>> measuring the aggregate network bandwidth, it is perfectly normal for
>> the bandwidth to scale linearly with the number of running rostopic bw
>> instances.
>>
>> A message lower down in the thread says that transferring an
>> uncompressed video stream transfers 8MBps. This seems to indicate to
>> me that there is no problem sending more than 8MBps over the network.
>>
>> Could you explain why you think that 4MBps is not sufficient for
>> sending the compressed stream, and what problems you are actually
>> seeing. From your report things are behaving exactly as they should.
>>
>> > A colleague of mine noticed something about wpa_supplicant_node having a
>> > wireless 11b max bandwidth limiter, is this what is limiting our
>> transfer
>> > rate? Does anyone know how to unlock the limit?
>>
>> This seems very unlikely to me, but where exactly did your colleague
>> see this limit?
>>
>> The reason this seems unlikely to me is that on the pr2
>> wpa_supplicant_node is not used. All wireless goes through the 610n
>> router in the robot, and hence the "wireless" traffic is not at all
>> wireless as far as the pr2 computers are concerned.
>>
>> Moreover, wpa_supplicant is only concerned with setting up the
>> security on the wireless connection. It does not do any packet
>> filtering. I have definitely used it at > 100 Mbps when network
>> conditions are good.
>>
>> Cheers,
>> Blaise
>>
>
>
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20110310/6f0c6404/attachment-0003.html>


More information about the ros-users mailing list