PipelineDB

PipelineDB was a PostgreSQL extension for high-performance time-series aggregation, built to support real-time reporting and analytics applications. Its central abstraction was the "continuous view": a SQL view that selects from a combination of streams and tables and is incrementally updated in real time as new data arrives, functioning as a very high-throughput materialized view that never requires manual refresh. Only aggregate output was stored in regular, queryable tables; raw time-series rows were discarded once read by the continuous views that needed them and were never written to disk, which made the system efficient for aggregation workloads but eliminated the ability to explore raw historical data.

Streams were exposed as PostgreSQL foreign tables managed by a "pipelinedb" foreign server, so events could be written using ordinary SQL INSERT statements and read with ordinary SELECT statements. Continuous queries emitted their own output streams — each output row for a continuous view carrying an old and a new tuple representing the incremental change — allowing queries to be chained into arbitrary networks of continuous SQL. Additional capabilities described in secondary sources include a HyperLogLog implementation for constant-space distinct counting, other probabilistic data structures, sliding windows, and stream-table joins.

The extension was written in C, derived from PostgreSQL, used a relational data model with a disk-oriented, row-based storage architecture, and exposed a SQL query interface. The final open-source version ran on 64-bit architectures and supported PostgreSQL 10.1–10.5 and 11.0; an earlier design description references PostgreSQL 9.4 compatibility. ZeroMQ was used for inter-process communication, and the build additionally required PostgreSQL development packages and Python dependencies for the test suite.

Founding story

PipelineDB was founded by Derek Nelson, who built the product and took it through Y Combinator's Winter 2014 batch. Nelson's prior experience is listed as including AdRoll, and after the acquisition he became a senior product manager at Confluent working on Apache Kafka and serverless Flink offerings. One profile refers to him as a co-founder, while Y Combinator's listing names him as the sole listed founder.

Business model

PipelineDB was distributed as an open-source PostgreSQL extension (the GitHub repository lists an Apache-2.0 license; a database catalog lists BSD and GPL v3). One catalog classifies the project types as both commercial and open source, and a retrospective article references an enterprise edition that was intended to address scalability limitations of the open-source product. No pricing or revenue details are available in the sources.

Not described in the sources beyond the classification of the project as both commercial and open source and a reference to an enterprise edition.

Traction

The GitHub repository shows approximately 2,700 stars, 243 forks, 97 watchers and 1,023 commits. Y Combinator lists a team size of 2 and a status of Acquired. DCVC lists PipelineDB as a portfolio company with a first investment in 2014 and records the outcome as an acquisition exit by Confluent (CFLT).

Latest developments

The most recent development recorded in the sources is the 2019 acquisition by Confluent and the subsequent shutdown of the open-source project, whose last commit is dated May 1, 2019. A database catalog entry accessed in June 2026 lists the project as abandoned with an end year of 2019 and notes that the previously listed project website resolves to unrelated content. The domain pipelinedb.com as fetched today serves unrelated online-gambling content and no longer represents the company. A retrospective article published in October 2025 positions PipelineDB as one of the earliest streaming database solutions, preceding ksqlDB, Materialize, RisingWave, and Timeplus's Proton.