Hello authors,

Thanks for writing up the draft. I have couple of questions/clarifications.


  1.
Section 1 mentions head-of-line blocking and connection migration but does not 
quantify or reference any measurement.


  1.
In Section 4.4.2, the draft states:

"For Statistics Report message (regardless of whether it can get AFI/SAFI or 
not), since it may be a periodic message with low load, and different 
statistics may have some correlation. So it SHOULD always be carried over the 
control channel to maintain better atomicity of all statistics."
I believe, the control channel is defined as a single bidirectional Stream 0 
for the entire BMPoQUIC connection, and there is only one control channel, 
shared across all peers.

If so, will not it reintroduces serialization bottlenecks? e.g. If a router has 
500 BGP peers each sending periodic Statistics Reports, all of them are 
serialized on Stream 0. A large burst of stats from many peers could back up 
the control channel, potentially delaying Peer Up/Down notifications and 
Initiation/Termination messages.

Additionally, the "low load" assumption may not hold at scale, since RFC 7854 
allows Stats messages to contain many counters and at scale across hundreds of 
peers, the aggregate volume may not be trivially small. With many WG drafts 
adding Stats Type, this will further increase counter volume in future.
Hence, a natural improvement would be to allow Stats to go separate channel, 
and optionally, go on Per-Peer common function channel.


-thanks,
Aseem
_______________________________________________
GROW mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to