Azure Scenario

Distributed Performance Testing with Azure Load Testing

Run JMeter from multiple regions, push metrics to App Insights, gate releases on p95 + error-rate criteria.

Architecture

Azure DevOps ─► AzureLoadTest@1 task
                       │
               Azure Load Testing
                       ├─► engines: 10 across 2 regions
                       └─► target: Container Apps API
                                  │
                                  ▼
               App Insights (server-side dependency map + p95)
                                  │
               pass criteria: p95<800ms, error_rate<1%, RPS>=2000
                                  │
               PASS ─► promote release   |   FAIL ─► block + open work item

Services used

Azure Load Testing Application Insights Azure Monitor Azure DevOps Pipelines Azure Blob Storage Azure Container Apps

Steps

1. Define test
Author JMeter `.jmx` or URL-based test; commit `config.yaml` with pass criteria and engine count.
2. Trigger from pipeline
`AzureLoadTest@1` task in Azure DevOps runs the test with 10 engines across 2 regions.
3. Correlate server-side
App Insights instruments the target; load-test run id is sent as a custom dimension on every request.
4. Gate
Pass criteria evaluated server-side; pipeline fails if p95 > 800ms, errors > 1%, or RPS < 2000.
5. Trend
Each run uploads raw results to Blob; Workbooks chart p95 over the last 30 runs.

Takeaways

Perf is a release gate, not a quarterly project.
Server-side correlation pinpoints which dependency causes p95 regressions.