pivot
GitHub Get started

Real-time analytics database, built on open data formats.

Pivot brings the speed and concurrency the AI era demands to the lakehouse. Run low-latency, high-concurrency queries on a single architecture that unifies real-time analytics and lakehouse workloads.

$ curl https://pivotlake.io | sh copy
REAL-TIME DATABASES LAKEHOUSES ClickHouse StarRocks Druid Pinot Databricks Snowflake Spark Trino pivot sub-second queries on open formats Fast, but runs on proprietary formats and local disks. Open, but queries take seconds and are expensive.

01 - FASTThe fastest query engine now runs on open formats.

Relative query time, lower is better.Why is Pivot fast? →

ClickBench
43 queries
pivot (Parquet)1.09×
ClickHouse (native)2.17×
DataFusion (Parquet)3.09×
ClickHouse (Parquet)3.61×
DuckDB (native)4.17×
DuckDB (Parquet)4.31×
Spark53.73×
TPC-H
SF1000
pivot (Parquet)1.00×
DuckDB (Parquet)2.16×
DuckDB (native)3.10×
ClickHouse (native)4.68×
DataFusion (Parquet)7.37×
ClickHouse (Parquet)27.14×
Spark43.59×
Star Schema Benchmark
SF1000
pivot (Parquet)1.00×
DuckDB (Parquet)2.58×
ClickHouse (native)3.36×
DuckDB (native)5.00×
DataFusion (Parquet)12.86×
Spark44.38×
ClickHouse (Parquet)120.51×

Benchmark harness and queries are in the repo - reproduce them yourself.

All runs: AWS c8gd.metal-48xl

02 - ARCHITECTUREBuild a lakehouse with Pivot, or query the one you already have.

Ingest, compact, and query data with Pivot. Already have Iceberg tables? Connect your catalog and query them in place.

Clients
Apps
BI tools
AI agents
Pivotserver · standalone
Ingest

Stream small changes or load large batches.

Compact

Continuously merge and reorder files so queries read less data.

Soft ordering →
Query

Run sub-second SQL queries at high concurrency.

Object storageIceberg · Delta Lake
Amazon S3
Google Cloud
Azure Blob

03 - PORTABLERun anywhere. Query the same Parquet.

Run Pivot as a shared server or locally for ad hoc queries and AI agents. Both access the same tables in object storage, with no extra copies to maintain.

Server mode

Connect applications, BI tools, and SQL clients to a shared Pivot server or cluster using the Postgres protocol.

Standalone mode

Run Pivot locally from your terminal or AI agent and query object storage directly, with no server to manage.

Clients
Apps
BI tools
psql
Terminal
pivot open s3://lake
└ opened 3 tables
AI agent
pivot open s3://lake
└ opened 3 tables
Postgres protocol
Pivot server
pivot server
└ listening on :5432
Shared tables in object storage
Same tables. Same source of truth.

04 - SCALABLEScale up, down, or to zero - almost instantly.

With object storage as its backing store and only state, Pivot can be scaled up, down, or to zero almost instantly.

Object storage
eventsDelta
ordersDelta
customersIceberg
instance 1
pivot server
serving
instance 2
pivot server
serving
instance 3
pivot server
serving
instance 4
pivot server
scaled to zero
instance 5
pivot server
scaled to zero
Read the docs Star on GitHub