#323 Bounceback should have a to & fro bandwidth stat

[Robert McMahon]

The --bounceback option should provide both a to and from bw statistic, particularly since the request and reply size don't have to match.

[Robert McMahon]

here's a proposed output

rjmcmahon@fedora:~/Code/bb/iperf2-code$ src/iperf -c 127.0.0.1 -e -i 1 -t 3 --bounceback=1 --bounceback-period 0 --bounceback-request 1000 --bounceback-reply 500 
------------------------------------------------------------
Client connecting to 127.0.0.1, TCP port 5001 with pid 63480 (1/0 flows/load)
Bounceback test (req/reply size =1000 Byte/ 500 Byte) (server hold req=0 usecs & tcp_quickack)
TCP congestion control using cubic
TOS defaults to 0x0 (dscp=0,ecn=0) and nodelay (Nagle off)
TCP window size: 2.50 MByte (default)
------------------------------------------------------------
[  1] local 127.0.0.1%lo port 47210 connected with 127.0.0.1 port 5001 (bb w/quickack req/reply/hold=1000/500/0) (icwnd/mss/irtt=162/16640/32) (ct=0.08 ms) on 2024-05-11 13:27:11.441673 (PDT)
[ ID] Interval        Transfer    Bandwidth         BB cnt=avg/min/max/stdev         Rtry  Cwnd(pkts)/RTT(var)   Tx/Rx bytes  RPS(avg)
[  1] 0.00-1.00 sec  63.2 MBytes   530 Mbits/sec    44171=0.022/0.016/0.389/0.003 ms    0 639K/(10)/13(0) us  Tx=42.1 MByte Rx=21.1 MByte  RPS=46006
[  1] 1.00-2.00 sec  45.3 MBytes   380 Mbits/sec    31671=0.030/0.022/0.086/0.007 ms    0 639K/(10)/19(0) us  Tx=30.2 MByte Rx=15.1 MByte  RPS=32967
[  1] 2.00-3.00 sec  38.0 MBytes   319 Mbits/sec    26581=0.036/0.031/0.077/0.003 ms    0 639K/(10)/19(1) us  Tx=25.3 MByte Rx=12.7 MByte  RPS=27715
[  1] 0.00-3.01 sec   147 MBytes   408 Mbits/sec    102424=0.028/0.016/0.389/0.008 ms    0   -(-)/-(-)    Tx=97.7 MByte Rx=48.8 MByte  RPS=35565
[  1] 0.00-3.01 sec  BB8(f)-PDF: bin(w=100us):cnt(102424)=1:102421,2:1,4:2 (5.00/95.00/99.7%=1/1/1,Outliers=0,obl/obu=0/0)
rjmcmahon@fedora:~/Code/bb/iperf2-code$ 

[Andreas]

Hi Robert, I came across this topic because it mentions the bounceback option in the title.

We frequently use iperf in our measurement applications, where we plot live throughput graphs over time. As a result, the input for our graphs is based on the interval-level output rather than the final output. In our applications, and likely in many others in the industry, it's essential to differentiate between upstream and downstream throughput. Therefore, it would be beneficial to have separate Transfer and Bandwidth values for Rx (receive) and Tx (transmit) per interval."

[Robert McMahon]

Hi Andreas,

Are you asking the server side output to be on the client side too? Currently the server side output is only on the server side and the client side output is on client side.