PancakeDB solves a problem that has stymied data engineering for a decade: making streaming data accessible to batch and offline analysis.

Parquet uses 194MB whereas PancakeDB's new format uses only 116MB.

Write like a stream, read like a ton of bricks.

With write response times of only 10ms and read throughput of millions of rows per second per connection, PancakeDB is unlike anything you've seen before. It leverages a new columnar format using 30-50% less network bandwidth and storage than .snappy.parquet.

A pancake sits atop a pile of money.

The simplicity data engineers crave.

Founded by a data engineer to solve this problem universally, PancakeDB lowers storage, compute, and engineering costs dramatically.

Use Cases

Company A is a new startup. They don't have any support for offline analytics yet, and that's starting to become a problem. They're hoping to start a simple but extensible data stack to do ad-hoc analysis.

Company A would use PancakeDB to:
  • make streaming data available for offline analysis with easy setup
  • get things right the first time
Company A before PancakeDB Company A after PancakeDB

Company B is a mid-size startup. To support their scheduled and ad-hoc analytics tools like Spark, Hive, and Presto, they take nightly snapshots of certain tables in their production database and upload them as Avro files in S3.

Company B would use PancakeDB to:
  • enable real-time data freshness instead of nightly
  • alleviate load on the production database
  • support much faster reads than from Avro
  • save engineering maintenance
Company B before PancakeDB Company B after PancakeDB

Company C is a large corporation. To support a wide range of analytics and machine learning use cases, they have services to ingest event data and processed streams as small Parquet files in a data lake, then compact those small files into large Parquet files.

Company C would use PancakeDB to:
  • save millions per year in storage and compute
  • support even faster reads than from Parquet
  • save engineering maintenance
Company C before PancakeDB Company C after PancakeDB