RE: Re: [psg.com #471] Expert review comments on draft-ie tf-ipcdn-doc sisevent-mib-03

First off, I am not aware of anyone actually using the bulk of the defined events at the moment. This could be for a number of reasons from the fact that a large percentage of these events should be for rare events but when backend services break or a large percent of your CM footprint is running firmware of poor interoperability, you don't really want to get a flood of events from CMs to tell you such. Oddly, for all the events defined, the only CM event I am aware of in active consideration is CM de-range (offline or flapping) which is kindly provided as an enterprise TRAP from most CMTS vendors but MSOs have been historically reluctant to enable it. If I am mistaken, I would obviously be interested to know otherwise.

Mind you, there is recognition of the desirability of the timeliness of notification with these events so I would expect usage of the defined events to increase on a per-task basis -- e.g load-balancing will drive usage of the DCC events; software upgrade applications may already consider upgrade events. 

So my argument on events being rare is conjecture, but based on the nature that the bulk of the events should only fire in anomalous/rare situations in a healthy system or in bulk in an unhealthy system (the case I ignored for the poor fit of the granular event model to such a case without an aggregating agent in the CMTS or as part of the NMS).


-----Original Message-----
From: Jean-Francois Mule [mailto:[email protected]] 
Sent: Tuesday, August 03, 2004 2:21 PM
To: Fred Oko; Nakanishi Greg-MGI8179; [email protected]; DOCSIS OSS Majordomo List
Subject: RE: [ipcdn] Re: [psg.com #471] Expert review comments on draft-ie tf-ipcdn-doc sisevent-mib-03

Fred Oko wrote:
> For point1, we are talking at most 100 extra bytes, and on average
> closer to 50 -- fairly insignificant in the grand scheme of things
> unless we would send a ton of these notifications which should not be
> common. But admittedly a large % inefficiency when considering the
> entire PDU w/o event text is not much more than 100 bytes.

Thank you for your input. Do you or anybody have any actual counts of the number of such messages we typically get at the node/network level in 1 hour?

Based on the issues raised for simple NMS apps, trap exploders and alike using event level to properly route those msgs, I would really like to have a sense of whether this is a theoritical argument or whether we better be concerned about the verbose msgs.

Jean-François