The purpose of model concurrency testing is to validate that the model's key user journeys will perform optimally at the customer’s expected level of concurrent user interaction. If testing identifies issues, there are actions that the implementation team can take to address and ensure optimal performance.
The implementation team must verify that model performance for a single user, or at very low concurrency (2-3 users), is satisfactory prior to model concurrency testing. This is because the impact of any model performance bottlenecks will increase significantly as the concurrent usage increases and the model concurrency test results will be very poor.
Overview
Engage the model concurrency team
Define user journeys and scenarios
Create a sanitized version of the model
Validate model configuration
Create test scripts
Run tests
Generate reports and communicate results
Optimize the model
Model delivery lifecycle
