How a billion-dollar e-commerce business built a performance testing framework from scratch

Challenges

This e-commerce company had experienced steady sales growth, but in 2020, online shopping demand surged dramatically. The company saw exponential traffic increases, especially during holiday seasons.

Since then, BOT traffic and web scrapers have become increasingly sophisticated, creating challenges in managing real user experience. Filtering all BOT traffic wasn’t an option, as it could impact legitimate customers. To combat this, the company must ensure it can handle multiples of its standard traffic to avoid performance issues even if it faces heavy BOT traffic.

Additionally, the company needed to build a performance testing framework from scratch to manage site traffic effectively and maintain its commitment to high-quality service.

Solution

With Gatling Enterprise, the company started load testing daily, with several simulations to replicate traffic spikes and production traffic. Based on the simulations, they changed the architecture of certain pages, making them more resilient and protecting them from large amounts of traffic.

Currently, the performance team is continuously doing CI/CD testing, including soak and stress tests, to maintain the website's stability and uninterpreted customer service.

"We did a POC and evaluated multiple tools. Gatling was chosen because it was developer friendly, could be hosted in AWS, allowed us to spin up unique injectors for each test, supported lots of concurrent tests running in parallel without impacting each other, and the Gatling Enterprise UI was easy to use."

Associate Director, Software Engineering

Results

By using Gatling Enterprise, this e-commerce leader has successfully scaled performance testing, mitigated risks associated with BOT traffic, and optimized website stability to handle massive spikes in traffic and sales.

• Millions of active customers with an average session duration of 08:19 minutes
• 450+ developers actively using Gatling Enterprise
• 500+ load tests and 100+ distinct simulations per month
• During peak seasons, up to 1,300+ load tests per month