About the service
I set up Trino (and classic Presto) as a unified SQL engine that reaches all your data at once. Instead of copying everything into one warehouse, I connect the sources you already have — object storage S3 with Hive or Iceberg tables, PostgreSQL, MySQL, ClickHouse, Kafka, Elasticsearch, Mongo — and give analysts one entry point where a single query can join a fact table from the lake with a reference that lives in the operational database. That's federation in practice, and that's where both the biggest wins and biggest mistakes hide.
I start with catalogs and the connector layer: I write clean connection configs, lay out schemas, pick the right table format, and tune the metastore or Glue so partitions are detected correctly. Then comes performance tuning. I size the coordinator and workers, set memory limits and spill, regulate split counts, push filters and aggregates into sources, and rewrite queries that scan far more than needed. On top, I set up security the way audits expect: authentication, TLS, column and row-level access control, query logging — so you can safely open the platform to the whole team.
You get documented, reproducible deployment, not a demo that falls apart under load. I can run Trino on bare metal, in Docker, or in Kubernetes, connect it to your BI tool and scheduler, and stay nearby to teach the team to read query plans and keep performance up. One SQL dialect over everything, no waiting for nightly exports.
— Connectors to S3/Hive/Iceberg, SQL and NoSQL sources, Kafka
— Performance tuning: memory, splits, filter and aggregate pushdown
— Security: TLS, authentication, column and row access, audit
— Deployment in Kubernetes or bare metal with monitoring and docs
I start with catalogs and the connector layer: I write clean connection configs, lay out schemas, pick the right table format, and tune the metastore or Glue so partitions are detected correctly. Then comes performance tuning. I size the coordinator and workers, set memory limits and spill, regulate split counts, push filters and aggregates into sources, and rewrite queries that scan far more than needed. On top, I set up security the way audits expect: authentication, TLS, column and row-level access control, query logging — so you can safely open the platform to the whole team.
You get documented, reproducible deployment, not a demo that falls apart under load. I can run Trino on bare metal, in Docker, or in Kubernetes, connect it to your BI tool and scheduler, and stay nearby to teach the team to read query plans and keep performance up. One SQL dialect over everything, no waiting for nightly exports.
— Connectors to S3/Hive/Iceberg, SQL and NoSQL sources, Kafka
— Performance tuning: memory, splits, filter and aggregate pushdown
— Security: TLS, authentication, column and row access, audit
— Deployment in Kubernetes or bare metal with monitoring and docs
