mailwal

QuestDB 10.1: less I/O for late data, and more analysis in SQL

QuestDB Team
QuestDB

Late data without the rewrite

We released QuestDB 10.1, the latest version of our open source database.

Live data often arrives out of order, especially from several feeds or parallel senders. Until now, QuestDB rewrote the whole affected partition to merge in the late rows. With merge-append partitions, it appends them and compacts in the background. Late data uses much less disk I/O and you can query it sooner. It's opt-in, with one setting (docs).

A QuestDB Web Console chart of rows written to disk per row ingested on the demo instance from 19 to 26 October. The ratio climbs every day to around 1,000 and resets, until 22 October, when it drops to a flat line near zero and stays there.

Rows written to disk per row ingested on our public demo. Over 400 on average before merge-append was turned on, under 10 since.

10.1 also adds SUBSAMPLE for charts, exact percentiles, and a history of QuestDB's own metrics in sys.metrics. Check the breaking changes before upgrading. The release post has all the details.

Where we were in October

We had a booth at two conferences. At the Space Software Summit in Westminster, Colorado, we talked telemetry with flight and ground software engineers. Thanks to the OpenC3 team for having us.

Two QuestDB team members at the QuestDB booth at the Space Software Summit, between two banners, with a live telemetry dashboard on a monitor and a LEGO NASA Artemis rocket set for the raffle

Steve and Nick at the Space Software Summit.

Two weeks later we were at STAC Summit in New York, where our CEO sat down with Brian Pomraning of Broadridge for a fireside chat on data infrastructure as markets move toward 23/5 trading.

Three people seated on stage at STAC Summit New York in front of an audience: a STAC moderator, Brian Pomraning of Broadridge speaking in the middle, and Nic Hourcard of QuestDB on the right

Brian Pomraning of Broadridge and our CEO Nicolas Hourcard at STAC Summit NYC.

More reading:

  • Which time-series join to use: ASOF, WINDOW, HORIZON, LT and SPLICE explained in plain language, each with a query that runs on the live demo. Which time-series join?
  • Live market data from Databento: stream every US equity quote and trade into QuestDB with one config file, and keep every row through restarts and failover. Ingesting live market data from Databento
  • Sending a timestamp in one bit: how QWP uses Gorilla delta-of-delta encoding, worked through on real ticks. QWP under the hood

Questions or feedback? Join our public channels.

Slack | Community Forum | LinkedIn | X

Copyright © 2026 QuestDB Technology Inc.

London Office: 8 Devonshire Square, London EC2M 4YJ, UK

New York Office: 575 Lexington Ave, New York, NY 10022, USA