Skip to main content
Version: 0.5

Tecton on Snowflake Deployment Model

Tecton’s deployment model gives you control over your data while benefiting from the agility of a SaaS solution. Tecton orchestrates best-in-class infrastructure that already exists in your stack, while abstracting its operations away, so you can focus on building and deploying ML applications.

Tecton operates out of a control plane and a data plane.

Architecture Diagram

The control plane, which lives in Tecton’s single tenant AWS account, is operated by Tecton to guarantee core services and manage metadata. Core services include orchestrating data pipelines and serving online features to applications running in production.

The data plane consists of your Snowflake account, where your data is stored and processed, and your AWS account, which may host whichever service you use to access your feature data for training and inference purposes. The data plane does not host any proprietary services from Tecton.

This architecture ensures data security and compliance while leaving operational overhead to Tecton. It enables you to receive constant updates and get access to new product features seamlessly. And it guarantees that Tecton engineers can independently maintain and resolve any urgent issues with your features, 24/7.

Each Tecton cluster is connected to a single data platform. If you are an existing Tecton user, Tecton on Snowflake will require a separate Tecton cluster. In the future, Tecton will allow multiple data platforms per cluster.

Transitioning to SaaS with self-hosted online store

After initial deployment, you may choose to host the online store in your own AWS account. With this option, Tecton’s Feature Server will access the online store in your account to retrieve and serve real-time features.

Architecture Diagram

Key Benefits of SaaS

Tecton recommends the SaaS deployment model with a managed online store over self-hosting the online store. SaaS is optimized for simplicity while allowing for future improvements at the same time. It offers three key benefits:

  • Simplicity: Tecton abstracts away the online serving infrastructure required to support your machine learning applications running in production.

  • Performance and Cost: Tecton makes continuous improvements to the managed online store, optimizing for performance and cost. Using the managed online store will allow you to benefit from these improvements automatically, without having to make any changes to your infrastructure.

  • Security: Data is always encrypted in flight and at rest using AWS owned keys, so you keep best-in-class security practices regardless of your deployment model. With SaaS, Tecton has reduced access to your AWS account.

Transitioning to a self-hosted online store after initial deployment may be the right choice for you if:

  • Your policies require additional monitoring and controls on the online store, above and beyond the monitoring Tecton offers out of the box.
  • You want to use Redis, instead of DynamoDB, as the online store.
  • Your feature data is subject to specific regulatory requirements, such as PII.