[ippm] Violated vs errored intervals - draft-mhmcsfh-ippm-pam-00

[Alexander L Clemm <[e-mail]>]

Hi,

I wanted to follow up on the draft concerning Precision Availability 
Metrics for SLO-governed end-to-end services 
(draft-mhmcsfh-ippm-pam-00), which Greg Mirsky presented at IETF 113.  
One of the motivations for the draft is to provide an agreed-upon set of 
metrics which can provide the basis for accounting for high-precision 
services that feature stringent service level (aka performance) 
guarantees.  We believe that monitoring such services, and indeed the 
ability to provide such services, will hinge on the ability to assess 
whether a service complies with its guarantees.  This will be an 
important enabler for mission-critical applications that depend on such 
guarantees and accordingly I hope that it will be adopted by the WG.

I wanted to pick up on one of the discussion items that are pointed out 
in the draft, for which we (i.e., the authors) would like to request 
feedback from the WG, concerning the terminology to use. The 
foundational metrics that are proposed are concerned with tracking the 
number of time intervals during which a service (generally, a flow) 
complies with its objectives, versus the ones when violations occur.  
The draft currently refers to "errored" intervals (in the sense that 
there is a performance error). However, it seems that the term 
"violated" intervals may in fact be more suitable and semantically 
accurate to use, in the sense that it is really SLO violations that 
occur (and what constitutes a violation depends on the particular 
objectives that govern the flow, as opposed to an "error" which 
constitutes, well, an error).  This would imply replacing the term 
"Errored Interval" (EI) with "Violated Interval" (VI) (including its 
derivatives).

Comments?

--- Alex