The Problem: To analyze 500GB of logs, I usually have to either:
- Wait 5-10s for Athena cold starts + pay $5/TB scanned.
- Download the data to my laptop (slow + egress costs).
- Upload it to a SaaS like Datadog/Snowflake (massive cost + privacy risk).
The Solution: Burner is a "Bring Your Own Compute" SQL engine.
Architecture: You run the Agent (Docker) on your own EC2 instance in the same region as your bucket. Its zero egress as the data never leaves your VPC. The Agent computes aggregations in-memory (using DataFusion/Arrow) and only streams the JSON results back to you. I integrated wasmtime so you can compile Rust logic (like PII scrubbing or CIDR matching) and run it inside the query engine at native speeds.
Benchmarks (1GB Parquet / 100M rows):
- Python (Pandas): ~45s (Crashed OOM on t3.micro)
- Burner: ~1.8s (Streaming batches, 256MB RAM limit)
Business Model: I'm trying a $19 Lifetime License (Beta) instead of a subscription.
Repo is here: https://github.com/burnersql/burner-agent
Email: hello@burnersql.com
Happy to answer questions! I hope this is useful. Enjoy folks. :)
siscia•2mo ago
However my approach would be to use duckdb and S3 over lambda.
Leaving many of the concerns to the infrastructure. Like basically no OOM. No need to manage servers.