The origin of Restate

While building Flink and working with its users, we often saw non-analytic use cases, which were fascinating, but not the best fit for Flink. Those applications included distributed transactional applications and more traditional RPC-based applications, often with critical correctness requirements, like user management, logistics, ecommerce, payment, etc.

For those applications, way too much time is spent on non-functional requirements. Whenever we interact with other services, APIs, or asynchronous tasks, we are building a complex web of retries, state machines, distributed locks, queues, and all sorts of complicated and error-prone synchronization techniques. Event-driven applications were supposed to fix this, but end up being complex to write, debug, and operate.

We saw an opportunity to improve this. Restate lets you write applications as familiar RPC services, but executes them with an event-driven foundation that gives you the scalability and almost magical resilience of stream processing.