-
Notifications
You must be signed in to change notification settings - Fork 4
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
TS26.512 - Clarification on the client behavior for the list of metrics which shall be reported #68
Comments
Yes, this looks ambiguous in TS 26.512. I tend to agree with @dsilhavy's preference that an empty To me this seems more useful than using an empty array to signal that metrics reporting is temporarily disabled, which would be the next most obvious semantic we could define instead. If agreeable, I think the first semantic above could usefully be clarified in the Description column of table 11.2.3.1-1. Any views to the contrary? |
If I remember correct, the asssumed way to disable "old-style" metrics reporting in 26.247 and 26.114 (i.e. not AF-based), is to send a new valid metrics configuration, but with an empty list of metrics. I don't think there is any statement saying this, but there is no explicit "on/off" parameter. I guess another way would be to set "reporting percentage" to 0 (zero), which in practice also would stop reporting. So we should maybe check a bit so that we have a consistent understanding on how this is done over our different metric-related specs. Or at least not add something which messes up things. Another issue is that some metrics require additional configuration information. For instance, the metric to measure buffer level needs to know the measurement interval. There are similar "extra parameters" also for 26.118 and VR metrics, and I'm not sure there are default values for all of these. |
Apologies, @dsilhavy: I accidentally omitted this issue from the agenda of yesterday's joint meeting with 3GPP SA4 MBS delegates. Looking at the M1 Metrics Reporting Provisioning API specified in clause 7.8, I notice that the MetricsReportingConfiguration.metrics array is an optional property which must, if present, contain at least one metric identifier. The specification adds:
So I think this is the intended behaviour when provisioning metrics reporting at reference point M1. Turning back to the Service Access Information derived from this at reference point M5 (clause 11.2.3.1), the equivalent array is always present, but is allowed to be empty. Following the logic through, I conclude that this corresponds to the case where the array is omitted at M1, with the semantic that the default set of metrics or (if that is not specified) all metrics are to be reported. I will therefore add a clarification to this effect in table 11.2.3.1:
|
As @gheikkila points out, reporting can be temporarily suspended by setting MetricsReportingConfiguration.samplePercentage to zero at reference point M1, so I think that's good enough as a crude on/off control. |
@dsilhavy: Regardless of the mechanism used to control metrics reporting, any change to the metrics reporting configuration only takes effect during the course of a media streaming session if the Media Session Handler re-requests the corresponding Service Access Information when it becomes stale. The Media Session Handler must honour the The Media Session Handler can easily determine if the Service Access Information has changed when its copy goes stale, for example by storing the The response is 304 Not Modified (with an empty response body) if the previously supplied resource representation is still fresh or 200 OK (including the updated resource representation in the response body) if it has changed. Only in the second case does the Media Session Handler need to reconfigure itself mid-session. In both responses, a new time-to-live is supplied in the (For reference, this behaviour is specified in TS 26.512 clause 4.7.2.3.) |
M1 Priovisioning (existing specification):
M5 Service Access Information (proposed clarification):
Of course, @dsilhavy, this begs the obvious question: what is the default set of QoE metrics for the default metrics reporting scheme in 5G Media Streaming (i.e., the one with scheme URI urn:3GPP:ns:PSS: DASH:QM10 specified in TS 26.247)? Clause 10.4 of TS 26.247 lists all the relevant DASH metrics, but doesn't specify a default set. The In the absence of any better information, I conclude for now that if the M1 MetricsReportingConfiguration.metrics array is omitted and the M5 clientMetricsReportingConfigurations[].metrics array is therefore empty, this means that all DASH metrics need to be reported by the Media Session Handler. Do you agree? |
Thanks for the detailed explanation @rjb1000
I will create a dedicated issue for this as it is currently not supported: 5G-MAG/rt-5gms-media-session-handler#30.
This sounds reasonable and corresponds to how metrics reporting has been implemented. |
Problem description
TS26.512 Table 11.2.3.1-1 defines the ServiceAccessInformation resource. The
clientMetricsReportingConfigurations.metrics
property defines the list of metrics which shall be reported, has atype
set toArray(String)
and acardinality?
of1..1
.I would like to understand if this means that the metrics to be reported by the Media Session Handler are always stated explicitly in the
metrics
array? Or is it allowed to leave themetrics
array empty which means that the Media Session Handler shall report all metrics that belong to the correspondingscheme
? In my opinion, the latter makes a lot of sense as the content provider probably does not want to explicitly enable all metrics for a specific scheme. Leaving the array empty can then be interpreted as "report all the metrics the media player supports that belong the target scheme".The text was updated successfully, but these errors were encountered: