Custom Bidder Built for Scalability
A leading cloud provider set out to build an OpenRTB bidder to explore performance and scalability limits.
Project overview
Our client’s goal was to design and build a bidder that utilised the OpenRTB protocol and determine what is possible from a performance and scalability standpoint.
Specifically, the goal was to build a bidder that could scale from hundreds
of thousands of bid requests per second to millions of bid requests per second.
Key challenges
High throughput
Addressing this problem required building a highly concurrent application, enabling horizontal scaling, and ensuring that both the data store and data stream could support the necessary throughput.
Low latency
Meeting the low-latency requirement was challenging and involved implementing non-blocking concurrency, efficient serialisation, and optimised I/O clients.
Additionally, overall latency is heavily influenced by the performance
of the sources and sinks we interact with for reading and writing data.
Results
- The bidder can process 150,000 requests per second (RPS), also
known as queries per second (QPS). - The bidder was able to keep latency below 10 milliseconds (ms) at the 95th percentile and below 30ms at the 99th percentile.
arrow_circle_rightCONTACT US
What to learn more about this project?
Get in contact with our team.
Radosław Kostecki
Business Unit Director