Description of problem:
By design PMD threads are continuously polling and for OVS-DPDK there is no way to determine the CPU usage for a DPDK application.

Use Cases

The service provider wants to use the CPU utilization of the VNF for capacity planning purposes to get the desired forwarding performance without over-provisioning CPU resources to the VNF
In a production environment VNF is dropping packets and Service Provider or Network Operator wants to understand the root cause for performance degradation (packet loss or increased latency) and rule out if CPU or memory allocation for the VNF is the bottleneck




This requirement is to support a mechanism / API which DPDK applications can use to expose CPU and memory usage to OVS and subsequently these stats can be used by the Prometheus exporter along with other telemetry data to the analytics application/client.  

Requirement	Details	MVP	Priority	EPIC
DPDP Application Vendor to implement DPDK API to expose telemetry data outside VM	
	Yes	1	

Red Hat must implement integration with DPDK API to populate OVS metrics data with the actual CPU usage for the VM so that the telemetry data can be consumed by the Prometheus exporter to the analytics client	
	Yes	1	

Validate the metrics exported by the Prometheus exporter with a dashboard client for the correctness of the telemetry data	
	Yes	1	



Refer to https://issues.redhat.com/browse/OSP-21835 for overall feature description. 

http://patches.dpdk.org/project/dpdk/patch/20221115153555.947526-1-rjarry@redhat.com/

http://patches.dpdk.org/project/dpdk/list/?series=26659

Note: This work requires documentation to be added to reflect the availability of the new metric and process for the VNF vendor to use the API to expose the metric on the DPDK side

Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:


Additional info: