Hacker News Reader

> latency trumps reliability

Let's say I don't care about the initial handshake latency, have disabled nagle and have no packet loss. Is there still a measurable latency difference between UDP and TCP? It would appear that there is (based on yours and other comments) - but I don't know why.

View on HN

No, assumign an already estabilished connection, TCP is just as fast if there's no packet loss, except for the tiny amount of time that the additional TCP stack logic takes on the CPU.

See eg here: https://cloud.google.com/blog/products/networking/using-netp...

You may have induced latency driven by window sizing/scaling/acknowledgement RTT as well.

If TCP says "you can't send more data yet because of congestion control" (like hitting the congestion window limit) then I guess you can categorise that as latency, depends on your viewpoint.

RTT of acknowledgements shouldn't normally enter into it, no? You can send while previously unacknowledged data is in flight (as long as congestion control state allows).

Primarily traffic cost, CDN services were much more important during the early days... and punching through NAT came with its own liabilities.

Reply on news.ycombinator.com