From Monolith to Workflows at Twilio

The Twilio messaging application used a monolithic architecture that was difficult to change. Their registration process, for instance, was a single state machine with 2,000 lines of code. This meant that adding a new step to the registration process required significant effort. Additionally, their system relied on many third-party integrations with rate limits, making it difficult to manage and troubleshoot errors.

Evaluation Process for Choosing a Workflow Solution

Twilio considered several workflow solutions, including Temporal, Apache Airflow, Netflix Conductor, and Amazon Step Functions. They chose Temporal because it met their criteria:

• Workflows as code: This was important because Twilio's workflows are complex with a lot of branching logic.
• Java support: Their existing monolith was written in Java.
• Minimal code overhaul: They didn't want to rewrite their entire codebase.
• Dynamic workflows: This was important because they needed workflows that could be updated even while running.
• Rate limiting: Temporal allows configuring the rate at which downstream APIs are called.
• Managed cloud solution: Twilio wanted to focus on building features and not managing infrastructure.

Benefits of Migrating to Temporal

Twilio observed several benefits after migrating to Temporal:

• Easier feature implementation: Temporal workflows are implemented as a sequence of steps, making it easier to understand and modify workflows.
• Improved developer productivity: The small learning curve of Temporal made it easy for developers to adopt.
• Faster development: Temporal workflows can be easily integrated into new features.
• Scale: Temporal allowed Twilio to scale their engineering teams.
• Easier onboarding: New team members were able to quickly learn Temporal and contribute to workflows.

Overall, migration to Temporal was a positive experience for Twilio. It streamlined their development process, improved developer productivity, and made their workflows more scalable and maintainable.

Replay Video

This summary was pulled from a talk that the Twilio team provided at Replay 2023.