[ippm] More Information on Question re: draft-morton-ippm-reporting-metrics-07

["Mcdysan, David E" <dave.mcdysan at verizon.com>]

Below is the document I had in mind when I made a comment in the meeting
today regarding long-term performance and how that relates to
availability. 

Section 7 of Y.1540 http://www.itu.int/rec/T-REC-Y.1540-200711-I/en

As I understand the subject draft, the implicit assumption is that the
service is always "available" as defined by the above reference.
Agreement on a time interval (e.g., 5 minutes) for when a service is
available (or not) and the PM (e.g., loss) threshold is an important
component of current commercial SLAs, where "long-term" is often on the
order of a month. The period of time Tav (e.g., 5 minutes), the criteria
and threshold (e.g., loss >25%) and minimum number of packets over which
to compute the threshold comparison (i.e., Mav) are things that could be
made a stronger recommendation to be less vague. Alternatively, a long
term performance measurement could state these parameters.

Including measurements of delay, delay variation, loss in the subject
draft that would be considered unavailable in the ITU-T definition cited
above would seem to potentially skew the distribution. 

I recommend that some mention of availability be made in section 6.2,
Long-Term Reporting Considerations, possibly after (or near) the
following statement:

"When the reporting interval contains many measurement intervals it is
possible to present the results as "metric A was less than or equal to
objective X during Y% of time."

Thanks,

Dave