Hi Dhruv,
Thanks for your review.
See inline
On 6/26/2022 4:03 PM, Dhruv Dhody wrote:
Hi WG,
I think this work is very useful. I have some comments -
Minor
- We need a reference or some discussion of what we mean by "intent"
before we jump into SAIN in the Introduction.
We could reuse the definition from draft-irtf-nmrg-ibn-concepts-definitions
Intent: A set of operational goals (that a network should meet)
and outcomes (that a network is supposed to deliver), defined in a
declarative manner without specifying how to achieve or implement
them.
- The statement "Such approaches are mainly suited for greenfield
deployments" about intent is not clear to me. Is the intent-based
approach (or SAIN) limited to new deployments? Maybe it is worth
expanding on or providing a reference.
We were trying to express that the brown field environments are move
complex from a intent point of view (multi-vendors, different
capabilities, different domains, silo organizations, etc.). I understand
your confusion regarding this paragraph.
We'll try to improve this paragraph. If still unclear, we propose to
remove it.
- It is not clear to me the various difference between the Expression
graph, assurance graph, and computational graph.If the expression
graph/computation is all about how the health score is to be
calculated, why is that not in the YANG (which only talks of assurance
graph)? Also, I don't understand the difference between subservice and
service expressions! If the subservice has dependencies, does it
become the same as the service expressions? The model YANG only has
subservices BTW!
Building the computational graph is an implementation choice. When
discussing implementation, we thought this would be useful, even if it's
out of scope of the draft.
- Does SAIN orchestrator has any say in what metric to collect and how
it impacts the health score? I understand in most cases it is to be
analyzed from the configuration but at the same time can that not be
influenced by the orchestrator?
The subservice is an abstraction that should take care of the metric.
Hence this is not the SAIN orchestrator job.
However, the composition of the subservice health scores could be part
of the SAIN orchestrator. However, this is an implementation choice.
- Some examples of how the health score might be calculated as an
integer between 0-100 would be nice in the example provided esp for a
case where it is something in middle like "60".
An interface down is 0
An interface losing packets is ... 50 or 40 or 60? basically, it depends
on the service impact.
So it's difficult to us to provide some guidance in this draft.
Nits
- Section 1, the closing parenthesis is missing -> (Section 3.3 of
[RFC8969]
- Expand on first use - L3VPN, dBM,
- Add reference for kubernetes, Openconfig
Thanks, Benoit
Thanks!
Dhruv
On Wed, Jun 8, 2022 at 3:29 PM Tianran Zhou
<zhoutianran=40huawei....@dmarc.ietf.org> wrote:
Hi WG,
This mail we start a two weeks working group last call for
draft-ietf-opsawg-service-assurance-architecture-03.
https://datatracker.ietf.org/doc/draft-ietf-opsawg-service-assurance-architecture/
Please send over your comments before June 22.
Please also indicate if you think this document is ready to progress.
Cheers,
Tianran, on behalf of chairs
_______________________________________________
OPSAWG mailing list
OPSAWG@ietf.org
https://www.ietf.org/mailman/listinfo/opsawg
_______________________________________________
OPSAWG mailing list
OPSAWG@ietf.org
https://www.ietf.org/mailman/listinfo/opsawg
_______________________________________________
OPSAWG mailing list
OPSAWG@ietf.org
https://www.ietf.org/mailman/listinfo/opsawg
