Evolution of Live Streaming Load Testing and Stability Assurance for an Online Education Platform

With rapid growth in user traffic after the 2020 pandemic, the online education platform needed a robust R&D stability system to support its live‑streaming services.

The performance testing journey is described as evolving through four eras: the "Stone Age" with basic tools, the "Bronze Age" with integrated micro‑services and richer scenario coverage, the "Electrification Age" after a demanding summer peak, and the aspirational "Information Age" aiming for automated, intelligent testing.

During the summer upgrade, the core services were migrated from PHP to Golang micro‑services, and the live‑streaming client was consolidated. Testing was divided into three stages: single‑interface, full‑scenario, and large‑session simulations.

Full‑scenario testing was refined by splitting a live class into 13 smaller interaction scenes, while large‑session simulations reproduced high‑concurrency teacher and student interactions to uncover bottlenecks.

Post‑summer analysis identified long data‑preparation cycles, insufficiently realistic test data, and fully manual test scripts as major pain points.

Improvements included updating 30 key scenarios, creating a "zero‑effort" data‑preparation API that clones completed session data into a fresh session, and leveraging the Conan traffic‑replay platform combined with PTS to automatically generate load‑test scripts from real‑world data.

Automation further progressed with scheduled weekly performance scripts that record core interface metrics, generate reports, and notify engineers, enabling rapid detection and fixing of regressions such as slow SQL queries.

The team also began classifying recurring performance issues to enable automated diagnosis, linking symptom patterns to root causes like slow queries or network problems.

Overall, the evolution demonstrates a shift from manual, fragmented testing toward an integrated, automated performance assurance pipeline that enhances coverage, efficiency, and reliability for the platform's live‑streaming services.