[nanomsg] Re: Broadcast protocol
- From: Martin Sustrik <sustrik@xxxxxxxxxx>
- To: nanomsg@xxxxxxxxxxxxx
- Date: Fri, 13 Sep 2013 03:10:20 +0200
On 12/09/13 18:17, Nico Williams wrote:
Indeed, generic reliable multicast is not possible: the limit is given
by the publisher's willingness to cache old messages. But if you set
that limit high enough it will just work well enough most of the time.
The typical problem with semi-reliable multicast is that the first thing
people do is to write busy loop sending messages to test the max throughput.
Obviously, they overload the tx buffer which then starts dropping messages.
One way to deal with that is implementing leaky bucket algorithm for
pub/sub combined with pushback to user when the buffers are full.
Martin
Other related posts:
