PERF-Expert: Novel Approach for Dynamically Forecasting the Performance of Cloud-Based Integrations with Zero Performance Tests

Organizations are increasingly adopting Cloud-based development to create services by integrating Cloud applications and Software as a Service (SaaS) APIs that meet performance standards set by Service Level Agreements (SLAs). However, optimizing such Cloud-based integrations through traditional load testing is challenging due to cost and Cloud application restrictions. Static analysis approaches like mean value analysis (MVA) can be used to minimize these difficulties. However, they require insights into invoked Cloud applications and make unrealistic assumptions. Alternatively, predictive performance modeling offers more accurate results, yet it necessitates the need for load testing the integration up to a certain level to obtain training data, which can be costly, especially when invoking third-party paid Cloud applications. To address these issues, PERF-Expert is introduced as a novel approach that provides dynamic and accurate performance feedback on Cloud-application integrations during development. PERF-Expert learns the performance characteristics of atomic Cloud applications within existing deployed integrations and combines them with mathematical models to determine the performance of new integrations without performance tests. PERF-Expert continuously monitors changes and provides dynamic feedback without human intervention. Compared to MVA, PERF-Expert is 34.75% more accurate for real-world deployments and can drill down to individual Cloud-application performance with high accuracy. PERF-Expert, backed up by a suitable Cloud observability platform, can be a powerful tool for developers and performance engineers. Illustrating its powerfulness, PERF-Expert is already integrated into a Cloud developer platform to provide development time performance feedback for Cloud-based integration.