throughput measured on usb device seems wrong

[jrsphoto]

I have an RTL-SDR usb sdr receiver, that I've been testing with usbtop to determin how much data its sending over the usb bus. I can use the command rtl_test -s 2.4e6 to tell it to sample at 2.4MHz bandwidth, which should put around 4800kB/s on the usb bus, but I'm only geting 1160kB/sec. In fact, if I change the sample rate to 3.4Mhz, which should peg the usb2 interface, It stays at 1160kB/sec.

I'm on ubuntu 16.04.3 LTS on an i5 processor with 16gb of ram.. Source compiled direct from github and compiled with no errors or warnings.

[aguinet]

Hello @jrsphoto !

Thanks for your tests! I was actually not very confident about the algorithm used for computing the bandwidth, and your results seem to confirm that! Do you have any suggestions/improvements on how the bandwidth measurement should be done? (that is, for instance, at which frequency should we compute it, using which sizes, etc...). Thanks!

[aguinet]

For information, all bandwidth computation related code is happening here: https://github.com/aguinet/usbtop/blob/master/src/usb_stats.cpp !

[mangodan2003]

Just like to add that I too think the results are inaccurate. I'm running tests between hosts with iperf using USB-Ethernet dongles.

The figures from iperf are:

Connecting to host 192.168.33.175, port 5201
[  4] local 192.168.33.1 port 49736 connected to 192.168.33.175 port 5201
[ ID] Interval           Transfer     Bandwidth       Retr  Cwnd
[  4]   0.00-1.00   sec  8.87 MBytes  74.4 Mbits/sec    0   50.9 KBytes       
[  4]   1.00-2.00   sec  9.10 MBytes  76.4 Mbits/sec    0   50.9 KBytes       
[  4]   2.00-3.00   sec  8.71 MBytes  73.0 Mbits/sec    0   52.3 KBytes       
[  4]   3.00-4.00   sec  8.64 MBytes  72.5 Mbits/sec    0   56.6 KBytes       
[  4]   4.00-5.00   sec  8.57 MBytes  71.9 Mbits/sec    0   80.6 KBytes       
[  4]   5.00-6.00   sec  8.51 MBytes  71.4 Mbits/sec    0   80.6 KBytes       
[  4]   6.00-7.00   sec  8.47 MBytes  71.0 Mbits/sec    0   80.6 KBytes       
[  4]   7.00-8.00   sec  8.62 MBytes  72.3 Mbits/sec    0   80.6 KBytes       
[  4]   8.00-9.00   sec  8.54 MBytes  71.6 Mbits/sec    0   80.6 KBytes       
[  4]   9.00-10.00  sec  8.55 MBytes  71.7 Mbits/sec    0    123 KBytes       
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth       Retr
[  4]   0.00-10.00  sec  86.6 MBytes  72.6 Mbits/sec    0             sender
[  4]   0.00-10.00  sec  86.4 MBytes  72.5 Mbits/sec                  receiver

whilst usbtop is saying around about :
Device ID 3 : 744.97 kb/s 11270.87 kb/s
which seems way of the mark, maybe a factor of 10?

Would expect to see something significantly higher than 72Mb/s taking usb overhead into account.

[Nable80]

11270.87 kB/s ~ 90.167 Mbits/s. Does it resolve your doubts?

[mangodan2003]

I would hazard a guess that its coincidence. The units displayed are kb/s - that's kilo bits per second, and USB throughput is normally measured in bits not bytes, presumably due to being a serial connection. I'll have a look at usb_stats.cpp and see if its a possibility though.

[Nable80]

I've looked at the code (src/usb_stats.cpp and src/console_output.cpp). All the calculations are performed in bytes and bytes per second, no conversion to/from bits is used, only the final conversion from bytes per second to kiB/s (division by 1024.0 in console_output.cpp).

[mangodan2003]

ah ok, not had a chance to look back at this yet but good to know. Thanks.

[dentys03]

Hi.

could you please clarify?
The output of usbtop is kb/s, kB/s or kiB/s ?

[aguinet]

It's in KiB/s (KiB as in https://en.wikipedia.org/wiki/Kibibyte)

[dentys03]

Thank you.
The output includes the usb overhead too?

[aguinet]

It does. Closing this issue. Do notbhesitate to re-open if you have any question!