On 12/5/17 10:12 AM, Dino Farinacci wrote:
draft-kouvelas-lisp-map-server-reliable-transport are looking for working-group
adoption of the draft. The proposed solution replaces the UDP based control
Augments. We cannot remove UDP since there is too much deployment. And TCP is
not always needed. And there is startup latency with the 3 way TCP handshake.
This will slow mobile handoffs.
Right, augment rather than replace. I believe there's no intention to
replace UDP. The draft actually relies on UDP registration to bootstrap
the TCP channel, and to ensure backward compatibility.
I also brought up at the WG that we should consider other transport layers that
are more secure than TCP by itself. Since your draft doesn’t spec how to use
TLS with TCP, QUIC may be a more lightweight, efficient and secure alternative.
QUIC would be an interesting option as well, probably worth of another
draft and possibly not limited to registration messages only.
I see this as a point optimization to the registration protocol, that
might be orthogonal to other transport-related mechanisms. In my
experience this has proved to be very effective in scalability of large
LISP deployments, especially with the increased volume of registration data.
Fabio
plane LISP communication between xTRs and Map-Servers with a reliable transport
session. This is done to eliminate the periodic messaging and improve scaling by
You can eliminate the periodic overhead by sending Map-Registers less often and
solicited Map-Notifies as acks. And you can test for map-server reachability
with RLOC-probes.
The only reason I bring this up is to disclose that we have existing mechanisms
to provide reliability.
And rather than have a completely new message parsing format, we could make
this proposal simpler by using existing messages that are sent over ANY
transport. I have made this comment before.
Note we did this in the PIM WG where existing messages are sent over TCP or
SCTP.
Dino
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp
