OK. What about making congestion state on TCP and SCTP more visible to
applications? I seem to recall at least a couple of RFEs open in that area.
One of the nice side-effects of this work could be making it possible
for pfiles(1) to show the congestion state of the open sockets in live
applications, which I think would go a long way towards making network
performance problems more observable.
The new socket options will give applications quite a bit of congestion
related information. The idea of showing additional state via pfiles
came up in a recent discussion, IIRC. We do take observability seriously
and are open to suggestions. It would probably be more efficient to talk
about this once we have a draft design - always easier to lead a
structured discussion off a frame of reference. I hope to have it
published in early December.
-Artem
_______________________________________________
networking-discuss mailing list
[email protected]
