I was just interested in whether the behaviour have been defined for those
who need early failure detection (systems with failover capabilities) and
are willing to pay for the additional bandwidth used (financial sector).
Why didn't you say so in the first place?
I believe that the common practice is to send two timestamped copies
of every packet which are routed over two distinct network paths, i.e.
not sharing circuits or routers, and then the receiver waits for both
copies to arrive. If the delay between the two identical packets is
too large, then the network is at risk, and you should assume the
worst, i.e. the information is out of date and should be discarded.
The timestamps are there to show you the variation in latency of the
first packet in the pair to arrive.
A couple of ways of sending two copies via two diverse paths are to
use MPLS where you can set up two LSPs over different topological
paths.or to use two different multicast trees which are routed
differently by your cooperating network provider.
Either SFTI or NYSE had published something about this which I
downloaded and read about 6 years ago. Can't remember all the details
but I remember that the intro talked about the transition from bisync
to IP. You really should be asking this kind of question elsewhere
where the financial types hang out. Latency is so important that even
the guys who are primarily software developers tend to know a lot
about how to do this.
--Michael Dillon
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf