Sum only seems to have a bug
Hi Bob,
I hope the highlighted numbers below are just a presentation issue.
It occurs once in a while.
Thanks,
Ram
iperf -s -e --trip-times --realtime -i 1 -P 10 --sum-only -Z cubic -p 5002
iperf -c 192.168.10.20 -e --trip-times --realtime -i 1 -P 10 --sum-only -Z cubic -p 5002 -t 10
------------------------------------------------------------
Client connecting to 192.168.10.20, TCP port 5002 with pid 8266 (10/0 flows/load)
Write buffer size: 131072 Byte
TCP congestion control set to cubic using cubic
TOS set to 0x0 (Nagle on)
TCP window size: 85.0 KByte (default)
Event based writes (pending queue watermark at 16384 bytes)
------------------------------------------------------------
[SUM-cnt] Interval Transfer Bandwidth Write/Err Rtry
[SUM-10] 0.00-1.00 sec 92.5 MBytes 776 Mbits/sec 740/0 161
[SUM-10] 1.00-2.00 sec 90.0 MBytes 755 Mbits/sec 720/0 40
[SUM-10] 2.00-3.00 sec 90.2 MBytes 757 Mbits/sec 722/0 43
[SUM-10] 3.00-4.00 sec 90.1 MBytes 756 Mbits/sec 721/0 40
[SUM-10] 4.00-5.00 sec 90.1 MBytes 756 Mbits/sec 721/0 41
[SUM-10] 5.00-6.00 sec 90.2 MBytes 757 Mbits/sec 722/0 41
[SUM-10] 6.00-7.00 sec 90.2 MBytes 757 Mbits/sec 722/0 40
[SUM-10] 7.00-8.00 sec 90.1 MBytes 756 Mbits/sec 721/0 42
[SUM-10] 8.00-9.00 sec 90.5 MBytes 759 Mbits/sec 724/0 40
[SUM-14] 9.00-10.00 sec 90.4 MBytes 758 Mbits/sec 723/0 40
[SUM-6] 0.00-10.01 sec 905 MBytes 758 Mbits/sec 7242/0 528
Need to separate interval thread cnt from the final thread counter. Code commit is here.
The other fixes didn't work. This one delays a dequeue if it's a final so the other intervals can go ahead of them. Commit is here. Needs testing.
Still got some issues