Scaling and benchmarking a critical message bus using a new indexing strategy

Aria, an internal messaging system processing terabytes of data daily, faced CPU overload when filtering messages for clients requesting only specific topic subsets during tip recovery operations. An intern optimized the system by implementing a new indexing strategy based on topic partitions rather than individual topics, achieving a 30% reduction in CPU usage on production workloads while maintaining system reliability. This solution prevents servers from reaching full CPU utilization and eliminates the need for temporary scaling fixes.