Postgres Summit US2026

PostgreSQL's logical decoding framework has emerged as a foundational primitive for Change Data Capture (CDC) in enterprise data architectures, enabling real-time propagation of database changes to heterogeneous downstream systems. However, the ecosystem built atop this framework — spanning multiple output plugins and a growing set of open-source consumer tools — presents practitioners with a fragmented landscape of overlapping capabilities, differing target support, and inconsistent handling of PostgreSQL-specific data semantics. This paper provides a structured comparative study of this ecosystem across output plugins and consumer tools.

On the plugin side, we compare pgoutput (PostgreSQL's native plugin), wal2json (JSON-based, broadly compatible), and decoderbufs (Protocol Buffers, used by Debezium), examining their output format characteristics, support for streaming large transactions, two-phase commit, sequence changes, partitioned tables, TOAST columns, and column/row filtering capabilities.

On the consumer side, we evaluate four prominent open-source tools: Debezium (Kafka Connect-based, targets Kafka and downstream Kafka consumers including Snowflake, BigQuery, and Elasticsearch via Kafka Connect sink connectors), PeerDB (PostgreSQL-native CDC engine to stream data from Postgres to Data Warehouses, Queues and Storage engines), Airbyte (EL(T) platform with a PostgreSQL CDC connector targeting various different destinations including data warehouses, databases, and lakes), and Apache Flink CDC (for real-time transformations enroute to analytical stores).