Re: [Dime] Resolution on action to discuss the need for Realm-Routed-Reports

[Dave Dolson <[e-mail]>]

It is difficult to judge this issue without knowing how the reported values are to be resolved when contradictory.

To send a request with both realm and host name, and having reported realm and host percentages, what should be done?

Does the client have to keep track of how much it has sent to each host, and also the realm as a whole?

This needs to be part of the spec, not left to vendor-specific implementation, because the overloaded realm/node needs a model of how the client will react.


The approach of having feedback for realm-only messages made it easier to understand what the client should do.

-Dave



From: Steve Donovan [mailto:[e-mail]]
Sent: Friday, March 21, 2014 05:05 PM
To: [e-mail] <[e-mail]>
Subject: Re: [Dime] Resolution on action to discuss the need for Realm-Routed-Reports

Ulrich,

The discussion should be captured in the minutes to the meeting.  I wasn't able to find them posted yet.

Jouni, Lionel, what is the status of the minutes for the meeting?

My reading of emails prior to the London meeting is different from yours.  I believe we had come to the conclusion that we needed host and realm (with the definition of realm as outlined below).  We were still discussing the need for Realm-Routed-Request reports.

Regards,

Steve

On 3/21/14 10:09 AM, Wiehe, Ulrich (NSN - DE/Munich) wrote:
Steve,

I don’t know what happend in London.
Can you please summarize the technical reasons that led to the London agreement.
E-mail discussions prior to London have clearly directed towards a report type that requests throttling of realm routed request messages (i.e. not containing a destination host) rather than a report type that requests throttling of messages routed towards a realm (no matter whether they contain a destination host or not).

Ulrich

From: DiME [mailto:[e-mail]] On Behalf Of ext Steve Donovan
Sent: Friday, March 21, 2014 3:33 PM
To: [e-mail]<mailto:[e-mail]>
Subject: [Dime] Resolution on action to discuss the need for Realm-Routed-Reports

All,

Ben and I took the action item to discuss the need for the Realm-Routed-Reports (RRR) report type.

As you may recall, the consensus coming out of the DIME WG meeting in London was to support two report types:

- Host -- Impacting requests with a Destination-Host AVP matching the host in the overload report (with the host implicitly determined from the Origin-Host AVP of the answer message carrying the overload report).

- Realm -- Impacting 100% of the requests with a Destination-Realm AVP matching the realm in the overload report (with the realm implicitly determine from the Origin-Realm of the answer message carrying the overload report).

The action Ben and I took was to come back with an opinion on whether RRR reports should also be supported.

My summary of the discussion is that we recommend to NOT include RRR reports in the current version of the base DOIC draft.

We still have some concerns with the granularity of control enabled by having just the two report types but the analysis of whether RRR reports are still needed can occur independent of the base DOIC draft.  If there is a determination that RRRs are needed in time to include in the base draft then it can be considered at that time.

Based on this, I propose the following

- Resolution to issue #23 is to remove RRR reports from the document.
- Resolution to issue #55 is to add Realm reports (actually to redefine them per the above definition).
- Resolution to issue #57 is that it no longer applies (as it deals with RRRs).

There is also need for text describing the interaction between host and the realm reports.  I don't expect we will have consensus on this wording prior to the -02 draft being submitted.  To this end, I'll open a new issue to deal with the need for wording on the interaction.

Regards,

Steve