Rollbar's overhaul of its Item Search backend aimed to address performance limitations and enhance search capabilities by rearchitecting the system to use specialized data stores. Previously reliant on MySQL for filtering and sorting, the architecture faced bottlenecks with complex queries and slow propagation of item relationship changes, leading to search delays. The new architecture assigns specific roles to MySQL, Clickhouse, and Elasticsearch, with MySQL handling static item-level filters, Clickhouse managing occurrence-level filtering and sorting with its columnar storage and parallel query execution, and Elasticsearch focusing on item title searches. This restructuring, alongside optimizations like near-real-time synchronization of merge data and simplified MySQL queries, resulted in significant performance gains, reducing query times from over 60 seconds to roughly 1-2 seconds and enabling real-time updates for merged items. The improvements also expanded search capabilities with more fields available for filtering and enhanced accuracy in time range filtering and sorting, leading to a faster, more reliable, and scalable search experience.