One further point:

Let's say Ivip is implemented and there are flaky losses of
connectivity with ECMP, LAG etc. - including especially new or
altered arrangements with different technical characteristics.

An Ivip-style reachability probing system (meaning it is completely
outside the Ivip system itself) could be updated to handle all these
potential problems.  For instance, from any one probing server to any
one ETR, the probing server could emit probe packets to the ETR which
are constructed to be hashed differently and so be sent via separate
paths.  With full knowledge of the potential multipath techniques
deployed between itself and any ETR, the probing server could
systematically try to have its probes go along every possible path.
Then it would be able to see the outages specifically, and so form an
estimate of general packet loss via the overall path, due to one or
more physical paths dropping packets.

Such sophistication and updating of probing protocols to meet the
challenges of the day would be out of the question in any ITR-based
system (LISP etc.) or any host-based system, because it is completely
impractical to upgrade all such ITRs or hosts to perform ever-more
complex probing.

 - Robin

_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to