How we achieve sub-8ms p99 ingest latency across 12 regions
A deep dive into our write path: from the producer SDK batching strategy all the way through the NVMe hot layer and replication acknowledgment protocol.
Post-mortem: 47-minute ingest degradation in us-east-1
On March 3rd, 2025, we experienced elevated p99 latency on ingest in our us-east-1 region. Here's what happened, what we missed, and what we've changed.
Introducing Axiom Storage: Apache Iceberg as a first-class citizen
Today we're launching the public beta of Axiom Storage — S3-compatible object storage with Iceberg tables built in. No more external metastore, no more ETL.
Building a columnar query engine that serves two masters: freshness and throughput
Our query engine needs to handle both fresh streaming data (seconds old) and historical cold-tier data (years old) with the same SQL interface. Here's how.
Migrating 2TB of Kafka topics to Axiom Streams with zero downtime
A step-by-step guide based on how one of our largest customers moved off self-managed Kafka over a weekend. Includes the MirrorMaker 2 config and rollback plan.
Why every engineer at Axiom does support rotation
We've had this policy since day one. Here's why we believe it makes better products, better engineers, and better customer relationships.