NCMP shall delete cmhandles at a rate of 50 cmhandles/second (need to confirm ?)  with cache disabled (assumes no dmi-plugin delays, assumes 180 modules per cmhandle with 95% module overlap between cmhandles)

Confirm

1. Environment Requirements for (CPS, NCMP and Postgres?!
2. Starting number of CM Handles, percentage deleted
3. Number of deletion requests (all in one?)
4. Assume deleted starts AFTER successful (all in Ready before use case starts)
5. CPS-Temporal Notification should be disabled for the NCMP Registry dataspace
6. CM Handle LCM Event enabled

Test suggestion, add 10,000 cm handles delete them al in one request (if possible) check timings

Confirm, CPS_temporal notifications are turned off, see https://jira.onap.org/browse/CPS-1361 (this might already improve the delete throughput)

• Add groovy test to verify the cm handles deletion performance? can extend existing read, write use cases to include deletion (ie. of 90% of the children created in the write/read test, see src/test/groovy/org/onap/cps/spi/performance/CpsToDataNodePerfTest.groovy)

Acceptance criteria (AC) :

• Agree between Stakeholder and CPS Team and document requirement on https://wiki.onap.org/display/DW/CPS-NCMP-I-01+Specification%3A+Performance+Requirements
• Demo CM handles are deleted (no longer in DB removed at a rate of 50 cmhandles/second. ie 10,000 Cm Handles deletion, after 200 second use SQL query to confirm all the cm handles are removed from the DB (and 20,000 CM Hande LCM notifications sent)
• Performance improvements documented in https://wiki.onap.org/display/DW/NCMP+De-Register+API+Performance 
• Get test result from Kubernetes E2E deployment by ESH teams up to 20,000
• Agree with stakeholder if performance is now sufficient or not (might need to renegotiated original delete performance requirement