A Performance Testing Pyramid

When we code a new class method, does its execution time fit within the product’s SLA for response time? When we provide an API for a real-time system, does it meet its performance budget? When we’re managing a data warehouse system, what is the latency between the ingestion component and the persistence server? When we’re planning to deploy a large-scale web application, how many servers do we need to support 1,000 requests per second and respond in 3 seconds? We’ve learned that the need for performance data spans from requirements engineering through system deployment: when code is written, components are integrated, and systems are deployed. We’ve also learned this requires a holistic approach of well-designed layers of tests to both identify issues quickly and gain confidence in our system as a whole. The performance testing pyramid is intended to be parallel to the functional testing pyramid, and measures performance strategically with micro-performance tests at unit and integration levels, and macro-performance tests at the system level. We’ll discuss our performance testing pyramid, its benefits, the trade-offs, and how such a portfolio of micro-performance and macro-performance tests may be targeted to meet demand for continuous performance data.