• Article by Zhou Sun: The recent article "HTAP is Dead" by Zhou Sun sparked a debate in the data community. HTAP aimed to integrate historical and online data, support flexible query methods and reduce business complexity.
• History of HTAP: OLTP and OLAP database workloads were one in the 1970s, separated a decade later, and HTAP tried to merge them in the 2010s. But practical challenges make dedicated, specialized systems more viable.
• Cloud impact: The cloud led to the move from tightly coupled warehouses to modular lakes. Data teams started assembling custom systems.
• Main players: HTAP was considered a requirement for emerging workloads years ago, with SingleStoreDB and TiDB as main players. Cloud data warehouses like Snowflake and BigQuery won by focusing on analytical processing and separating storage from compute.
• Current approach: Data teams are assembling custom systems using "best-in-class" components like OLTP systems, stream processors, Iceberg, query engines and real-time systems.
• Debate in the community: The article sparked a debate. Peter Zaitsev summarized that there is no "one size fits all" and HTAP makes sense as a feature but not a name. Many data engineers now reconsider the HTAP model, with PostgreSQL's success illustrating the shift.