Re: [Teas] I-D Action: draft-ietf-teas-actn-framework-02.txt

[Stewart Bryant <[e-mail]>]

Something that seems to be missing from this work are the concepts of 
isolation and latency
which seem to be requirements for services of the type described in this 
document.

We latency has two components, the physical latency of the path due to 
distance, and
latency due to waiting for other packets. Clearly optimizing the paths 
traffic takes
is the key to addressing the delay to physical path length, and 
considering congestion
effects, bandwidth management and the choice of bearer technology impact 
the second.

I did not see any mention of latency in the abstract interface to the 
SDN system and
yet to address that emergent requirement it needs to be there.

Secondly I understand that isolation is an emergent requirement. 
Isolation can
take many forms. The most common is the need to avoid traffic leakage, 
but it
can also include the need to avoid control plane interactions whereby 
the demands
of one service impact another, and data plane interaction where a heavy 
demand
by one virtual service spills over into latency effects in another, at 
least in
classic network layer technologies such as MPLS-Tx.

- Stewart