Commence deprecation of suppress_redundant

[earies]

Following up on the long standing PR: #114

This PR is to request for comment and implementation status as to the overall
deprecation of suppress_redundant within gNMI subscriptions.

Should this change be accepted, the gNMI specification will be updated as well
as relevant .proto and generated stubs.

[ncorran]

Happy to see this be removed. Cisco-XR dont support the option (and don't have any plans to support it in future)

[jsterne]

Happy to see this removed on the Nokia side.

[raghubk]

@dplore

As far as deprecating existing gNMI specification, we should understand is what dependencies or use cases exist regarding suppress_redundant and heartbeat_interval. It would be good to hear from the community if these are commonly implemented and in use features.

We at IP Infusion have suppress_redundant feature ready for shipping. Yes, implementation of this is involved, but worth it since the feature is very useful as it reduces the amount of data that is pushed to the collector significantly in most use-cases.

@earies
I did not see any good reason in this thread that warrants deleting this feature from the gnmi spec.

[raghubk]

@dplore

As far as deprecating existing gNMI specification, we should understand is what dependencies or use cases exist regarding suppress_redundant and heartbeat_interval. It would be good to hear from the community if these are commonly implemented and in use features.

We at IP Infusion have suppress_redundant feature ready for shipping. Yes, implementation of this is involved, but worth it since the feature is very useful as it reduces the amount of data that is pushed to the collector significantly in most use-cases.

@earies I did not see any good reason in this thread that warrants deleting this feature from the gnmi spec.

If needed we can discuss this in the next openconfig forum meeting. Thanks.
@earies @dplore @robshakir

[santanukar2000]

If we deprecate suppress redundant, how do you plan to implement the 'dynamic value' paths like counters, temp, cpu, ram and similar sensors. If its just sampling, its not optimal as we can reduce by using suppress redundant as there can be long periods that these sensors remain same and then change. If its on-change, then we are actually not doing on change, but suppress redundant internally with a small interval, as its practically not possible to monitor these at very fine granularity without impacting the system performance.

[dplore]

@dplore

As far as deprecating existing gNMI specification, we should understand is what dependencies or use cases exist regarding suppress_redundant and heartbeat_interval. It would be good to hear from the community if these are commonly implemented and in use features.

We at IP Infusion have suppress_redundant feature ready for shipping. Yes, implementation of this is involved, but worth it since the feature is very useful as it reduces the amount of data that is pushed to the collector significantly in most use-cases.

@earies I did not see any good reason in this thread that warrants deleting this feature from the gnmi spec.

Hi @raghubk , thank you for your comment. I did a quick search of IP Infusion documentation and I don't see any reference to configure a suppression of redundant values for telemetry. Can you show us evidence that this is a supported feature?

(I looked here in your release SP 6.6 for streaming telemetry commands)

[dplore]

If we deprecate suppress redundant, how do you plan to implement the 'dynamic value' paths like counters, temp, cpu, ram and similar sensors. If its just sampling, its not optimal as we can reduce by using suppress redundant as there can be long periods that these sensors remain same and then change. If its on-change, then we are actually not doing on change, but suppress redundant internally with a small interval, as its practically not possible to monitor these at very fine granularity without impacting the system performance.

Hi @santanukar2000, we are asking for evidence that this feature is implemented and even more importantly, in use in production networks. I agree that the feature is theoretically useful. But if there are no implementations and therefore no operational usage after so many years of being present in the gnmi specification, it is suggested by the community that it be deprecated.

[raghubk]

@dplore

As far as deprecating existing gNMI specification, we should understand is what dependencies or use cases exist regarding suppress_redundant and heartbeat_interval. It would be good to hear from the community if these are commonly implemented and in use features.

We at IP Infusion have suppress_redundant feature ready for shipping. Yes, implementation of this is involved, but worth it since the feature is very useful as it reduces the amount of data that is pushed to the collector significantly in most use-cases.
@earies I did not see any good reason in this thread that warrants deleting this feature from the gnmi spec.

Hi @raghubk , thank you for your comment. I did a quick search of IP Infusion documentation and I don't see any reference to configure a suppression of redundant values for telemetry. Can you show us evidence that this is a supported feature?

(I looked here in your release SP 6.6 for streaming telemetry commands)

@dplore
Hi Darren, feature is ready from development point of view. Is in QA phase and would be part of the upcoming release, slated for mid 2026. This is a feature one of the large data center customer shown interest in deploying in 2027. HTH.

The doc you referenced is earlier one and the upcoming release doc is not yet live.

[dplore]

@dplore

As far as deprecating existing gNMI specification, we should understand is what dependencies or use cases exist regarding suppress_redundant and heartbeat_interval. It would be good to hear from the community if these are commonly implemented and in use features.

We at IP Infusion have suppress_redundant feature ready for shipping. Yes, implementation of this is involved, but worth it since the feature is very useful as it reduces the amount of data that is pushed to the collector significantly in most use-cases.
@earies I did not see any good reason in this thread that warrants deleting this feature from the gnmi spec.

Hi @raghubk , thank you for your comment. I did a quick search of IP Infusion documentation and I don't see any reference to configure a suppression of redundant values for telemetry. Can you show us evidence that this is a supported feature?
(I looked here in your release SP 6.6 for streaming telemetry commands)

@dplore Hi Darren, feature is ready from development point of view. Is in QA phase and would be part of the upcoming release, slated for mid 2026. This is a feature one of the large data center customer shown interest in deploying in 2027. HTH.

The doc you referenced is earlier one and the upcoming release doc is not yet live.

Thank you for this! Please request your customer to engage with the OpenConfig community to voice their support and operational use case for this feature. Ideally they would identify themselves and comment on this issue.