Currently AMQ brokers doesn't have proper performance regression tests nor any baseline for each specific version if not the one of the opiske's Maestro tool or our version of https://github.com/softwaremill/mqperf (thanks to the effort of dlenoch@redhat.com).

The former is not maintained from long time and won't play nice with the RH performance team load generator (https://github.com/Hyperfoil/Hyperfoil), while the latter is focused on the test case for the https://softwaremill.com/mqperf/ blog post.
There's some value to maintain a version of the mqperf benchmark in our Ci for public exposure, but, both from an engineering perspective and for customer engagement we need a proper reproducible set of performance tests that we could share with customers and use to check regression before any broker release or significant changes.

This epic is about:

• implement tools to compose performance test use cases/scenarios: these "building blocks" could be used by engineer/community users to run simple tests in their local environment too eg in the form of existing Artemis CLI commands
• run the performance test scenarios on the RH Performance Team CI (or on our Messaging QE CI)
• collect results and telemetry to both capture regressions and share it with users/customers

The 2 client protocols under tests are AMQP and CORE and we're interested first in building simple scenarios until testing more complex one, matching specific customers use cases.

Simple scenario involves:

• non-durable, durable messages
• transactions
• paging
• large messages
• multiple/single producers/consumers/subscribers
• different size of messages
• response under load ie target throughput benchmarks and all-out benchmarks
• message load balancing tests (multi broker)
• replication (multi broker)
• mirroring (multi broker)

Candidate customer scenarios:

• send/receive durable messages while paging
• multi protocol tests eg producer with protocol X, consume with protocol Y
• mqtt scalability tests
• ...