[Spread-users] High latency in WAN

Stanisław Trytek tryteks at pit.edu.pl
Thu Apr 6 07:51:40 EDT 2006


I work with 3.17.3 spread version, I am going to switch to 4.0.0 version
in a few months.

Number of segments is 5 (one demon per one segment currently). Capacity of
test WAN network is about 36 - 62kb,
One of demons works in the part of the network where the link is being
lost very often and capacity can be even smaller than 36kb.
The network load is about 1 message per second, it is not a lot.
It leads of course to NETWORK state and what interesting - a few dozens of
seconds delay in message delivery. It means that messages were not being
delivered for let's say 100 seconds.
My qustions are:
Why is that? :-) I would know the answer knowing answers to the following

How WAN protocol (as far as I know it is still RING) deals with such
unstable links?

Where can I find how exactly that WAN protocol works (any documents)?

What exactly happens with connections among demons when there is a  
NETWORK state in WAN? How ring is formed in WAN?

Are messages (among segments) being sent over UDP/unicast? in ring? or  

Is HOP protocol implemented in 4.0.0 (some years ago I read about so
called "Spread 4", "research-wise", which was to support wide area  

I would be gratefull for any suggestions.

Stanislaw Trytek

More information about the Spread-users mailing list