Why ad platforms run six databases to answer one question

What this post added

This post analyzes the architectural challenges of ad platforms that lead to data sprawl, where multiple databases are used for different functions, resulting in increased operational costs and complexity. It highlights the shift from traditional batch OLAP and individual OLTP systems to the need for real-time HTAP systems. The post details how this data sprawl arises from specialized databases for auction lookups, advertiser dashboards, audience filtering, and caching, leading to redundant data copies and synchronization issues. It argues that the traditional split between batch and real-time processing incurs significant costs due to data migration, latency, and operational overhead. The post introduces Hybrid Transactional/Analytical Processing (HTAP) as a solution, enabling both high-speed writes and analytical reads within a single engine, thus consolidating diverse workloads and providing a unified freshness profile and single point of access.