I use Starburst as a cost-efficient hosted option for Trino for data integration and ad-hoc analysis across a broad range of data sources. It is surprisingly useful to query SQL Server, a Google Sheet, data in a blob store, and persist it in Postgres for downstream consumption.
In addition, the Galaxy platform features such as scheduling jobs, offering a data catalog, easy permission and access control management, and the strong technical support from Starburst make it a breeze to use compared to something like Athena.
I have removed data silos, sped up my pipelines (three to five times the speed of Redshift on a per-cost basis), and now have a single point of entry with consistent SQL semantics to all of my data systems.
Query federation coupled with excellent performance is the best feature by far. A consistent interface to all my data systems and a friendly UI that supports data personas from Analyst to Architect and everyone in between is extremely valuable.
As a hosted option, I wish I had more control over the cluster configuration, specifically regarding some of the more advanced options. Trino is extremely flexible and powerful, but some of this functionality is gated on the Galaxy platform.
Most users and admins will never need these features, but on occasion I have encountered issues that could have been resolved by a configuration change in five minutes rather than redesigning a data product. That said, I have a high degree of expertise with the tool, and this is more of a quibble than a major issue.
I have used the solution for four years.
If I have the choice of tooling for managing and interacting with data systems, I always choose Trino first and Starburst Galaxy if I am responsible for managing the deployment. My current team deployed on Redshift before I joined, and the first and best architectural choice I made was to migrate to Galaxy.
You pay for cluster uptime. It is important to be aggressive about autoscaling, as a single worker will get you a long way. I recommend never connecting a BI tool to your Galaxy cluster. Instead, write the data to Postgres or a hot database and serve it from there so you don't pay for expensive uptime to serve dashboards.
Having a good amount of expertise in the domain, I knew that Galaxy was the right choice for quick deployment. Having managed data at scale (hundreds of terabytes) in the past, I know Trino will get the job done without a lot of hassle.
Athena specifically has two major issues. First, connectors are restricted on write functionality and are more difficult to configure. Not being able to write through connectors is a deal breaker. Second, if you scale out enough, you will encounter issues due to Athena's shared tenancy model and then need to migrate to Trino eventually. It is better to save yourself the hassle.
If you are unsure about the service, try the free trial. You can be up to speed with your existing systems in half an hour.
