Load testing distributed web services - George Malamidis, loveholidays

January 2024 saw loveholidays' busiest period to date, with loveholidays.com serving 8,050 requests per second. Not only were our systems stable during this time, they also maintained strict performance expectations at a record low infrastructure cost.

This is the result of continued focus on performance and scalability. Key has been ensuring our systems undergo daily load testing to address any changes that degrade performance ahead of time.

This talk covers the design and implementation of the system we created to automate this, including open sourcing one of the underlying tools written in Go (https://github.com/loveholidays/ripley) and about how replaying real production traffic at multiples of the recorded rate offers insight into how our systems operate at scale.