Announcing the Community Edition of Varada Trino Connector

We are extremely proud to launch the community edition of our Trino connector, delivering instant 10x-100x performance and concurrency uplift on your existing Trino clusters.

This new community edition of Varada’s connector for Trino was born in the community. It’s free 🤩 (for clusters of up to 4 nodes), and is a part of our ongoing commitment and contribution to the Trino community.

The connector offers instant performance and concurrency impact with Varada’s autonomous indexing technology. Everything else remains as-is:

• No query rewrite needed
• No additional footprint to your existing architecture
• Continue to manage your clusters as-is: UI, endpoints, scaling capabilities, etc.
• Security policies, data governance policies also remain as is and in full control
• No call home features

To highlight the impact we run a 4-query benchmarking analysis against Trino.

Varada vs. Trino: Performance (seconds)

Varada vs. Trino: Cost per Query

As expected, indexing works best for selective queries (needle in a haystack) and will deliver dramatic impact as queries become more and more selective. But you can start to see the impact even on full scan queries, which in this case had a 35% reduction in elapsed time.

Deploying Varada Trino Connector

The connector works out-of-the-box and the deployment process is easy and quick (up to 10 min).

After making sure you are using SSD-enabled machines (i3.4xlarge, r5d.4xlarge, r5dn.4xlarge), simply run the script on each worker and the coordinator.

Following the installation, which should take a few minutes, queries that uses Hive or Iceberg can now use Varada and enjoy the impact of dynamic indexing and smart cache management. You can continue using your old Hive connector and enable users to decide which path to use according to workload requirements.

A few more comments:

• The connector is available for Trino 362 and 364 (more versions coming soon!)
• After changing your cluster to run on SSD-attached machines (i3.4xlarge, r5d.4xlarge, r5dn.4xlarge), please make sure SSDs are restricted to usage by Varada only
• The connector doesn’t require any additional cluster footprint; you may actually realize very quickly that you can reduce your compute resources due to indexes added by Varada

Are you ready to get started?
Download the Community Edition of the Trino Connector