Summary of "Workflows, a new abstraction for distributed systems" by Dominik Tornow (Strange Loop 2022)

00:00:00 - 00:35:00

This video discusses workflows, a new abstraction for distributed systems that allows for transparent failure detection and mitigation. Workflows are implemented via subroutines, which are transparent to the core routine itself, achieving one's semantics. This video also provides a code example.

• 00:00:00 Dominik Tornow discusses the difference between a proof of concept and a production system, noting that a production system needs to guarantee correctness even in the presence of failure. He provides an example of a production system, describing how it guarantees correctness even in the presence of failure by ensuring completeness in case of a failure.
• 00:05:00 Dominik Tornow explains the benefits of using workflows in distributed systems, which can provide developers with a fail-tolerant experience while avoiding any visible failure detection or mitigation steps.
• 00:10:00 This video discusses abstractions, and how they are fundamental to computer programming. Abstractions allow for the manipulation and naming of compound elements, and can change the domain of discourse. They can also transform entire worlds.
• 00:15:00 Dominik Tornow explains workflows, a new abstraction for distributed systems. Workflows allow for the creation of durable executions, which are sequences of steps that do not have an effect on the state of the world. Workflows are implemented as a sequence of external steps that drive the world forward.
• 00:20:00 Dominik Tornow introduces "workflows," a new abstraction for distributed systems that automates the mapping of durable executions to volatile executions. Workflows are implemented in a language of choice, and they provide strong execution guarantees.
• 00:25:00 This YouTube video explains Workflows, a new abstraction for distributed systems which allows for transparent failure detection and mitigation. Workflows are implemented via subroutines, which are transparent to the core routine itself, achieving one's semantics. This video also provides a code example.
• 00:30:00 Dominik Tornow discusses workflows, an abstraction for distributed systems, which allows for durable execution. He compares the mechanics of a durable execution platform to a simple example.
• 00:35:00 Dominik Tornow discusses the boundaries of failure tolerance and failure transparency in distributed systems, explaining that the process itself (or orchestration part) must be deterministic, while activities (tasks) can be non-deterministic. He also mentions logging as a tool for achieving this.