detect slow send buffer draining pace using TCP_INFO

[totaam]

See When TCP sockets refuse to die:

An example of checking the draining pace:

while True:
    notsent1 = get_tcp_info(c).tcpi_notsent_bytes
    notsent1_ts = time.time()
    ...
    poll.poll(POLL_PERIOD)
    ...
    notsent2 = get_tcp_info(c).tcpi_notsent_bytes
    notsent2_ts = time.time()
    pace_in_bytes_per_second = (notsent1 - notsent2) / (notsent2_ts - notsent1_ts)
    if pace_in_bytes_per_second > 12000:
        # pace is above effective rate of 96Kbps, ok!
    else:
        # socket is too slow...

Using TCP_INFO.

See also #2420

[totaam]

See also ticket:

• #2369#comment:3
• shadow/seamless debugging latency #2617
• Cluttered network doesn't seem to get detected quickly enough for batch delay to compensate. #2812
• catch limited bandwidth issues sooner #999

[totaam]

Based on How to get amount of non-ACK-ed TCP data for the socket?, Is it possible to read the number of bytes buffered for a TCP write socket? and socket.py (Wrapper module for _socket, providing some additional facilities): r27690 adds TCP_INFO to xpra info.

See also : Congestion detection by observing socket queues.

Next step: figure out how to detect congestion faster using this new data.

[totaam]

See also #3001

[totaam]

Excellent read from cloudflare: Optimizing TCP for high WAN throughput while preserving low latency