Re: [pm-dir] PM-DIR Review for draft-ietf-nvo3-framework-06.txt

["LASSERRE, MARC (MARC)" <[e-mail]>]

Hi Dan,

Thanks for your feedback.
See my comments below.

Marc

________________________________
From: nvo3 [mailto:[e-mail]] On Behalf Of Romascanu, Dan (Dan)
Sent: Wednesday, May 28, 2014 2:56 PM
To: [e-mail]
Cc: Benoit Claise; [e-mail]
Subject: [nvo3] PM-DIR Review for draft-ietf-nvo3-framework-06.txt

This is the PM-DIR review for draft-ietf-nvo3-framework-06.txt. I am the assigned PM-DIR reviewer for this I-D. This review refers only to performance metrics aspects

This I-D defines a framework for Network Virtualization Overlays (NVO3) and a reference model along with logical components required to design a NVO3 solution.


The I-D does not define performance metrics, so a RFC 6390 review does not apply.

Performance metrics are mentioned in one place in the document, in section 5.2.6 which deals with the interaction between overlays and underlays and the need to exchange information about performance in order to ensure that the overlays requirements can be met by the underlay paths. The metrics that are mentioned as examples belong to the IP metrics category (throughput, delay, loss, jitter). Mentioning the IPPM framework and even the relevant RFCs would have helped, especially as there is a need for standard and stable definitions and methods of measurement in order to allow for an exchange of information between layers.
Agreed. I suggest adding a reference to RFC2330 in the last sentence of section 4.2.6:

 "such as defined in [RFC2330]."



'Performance' is mentioned a few more times in the document, with no reference to performance metrics. For example in section 3.3 there is talk about 'adequate performance to VM applications'.  It is not clear what this exactly means - what are the metrics for measuring performance of VM applications and how are they determined as 'adequate' (or not).

Would "the expected Quality of Service" instead of "adequate performance" be acceptable?

A further reference to RFC6390 could also be provided.



In the other instances, the word 'performance"  is used in its general meaning.



Clarifying these issues would be very useful IMO, because the overall functionality of the overlays networks depends among other on the sufficient allocation of resources in the underlays. The metrics and methods of measuring performance need to be clearly articulated for this purpose.



Regards,



Dan