orchestrata_ultimate.kernel.performance

Reference › orchestrata_ultimate › kernel › performance

Performance and scalability subsystem (decomposed from performance_manager).

The performance and scalability toolbox of the kernel package, SPEC-001 story US-010. It measures how the kernel performs, protects it from overload, spreads work across nodes and tests how it fails. It was split out of a single performance_manager module, which remains as the shim performance_manager. Inside the package, only the facade imports the other modules; the shim is the only outside importer of the package initialiser.

The parts, by role:

• Facade: manager holds PerformanceManager, which creates and coordinates the rest.
• Measurement: metrics has the latency, throughput, memory and uptime trackers; sla has the targets and thresholds they are compared with.
• Protection: resources has the cache, connection pool, batch processor, ring buffer and backpressure controller.
• Distribution: cluster has the consistent hash ring, leader election and cluster manager; scaling has the autoscaler.
• Testing: chaos injects latency, errors and timeouts.
• Support: _common and init.

Callers record measurements into PerformanceManager; it feeds the trackers, and update_load moves the backpressure controller between states, which should_accept_request then uses to shed low-priority requests. evaluate_scaling turns load into a scale decision, and check_sla_compliance compares measurements with targets. The package has no process entry point. Start reading at manager. Nothing in the message-routing code of router or dispatcher imports this package; the related host-level sampler is resource_monitor.

9 modules, 2,666 lines of code.

Modules

Dependency diagram