Distributed Async Await • A new programming model for the cloud - Dominik Tornow - NDC Oslo 2024

Distributed async/await formalizes concurrent and distributed programming structures while preserving developer experience through functions and durable promises

Modern distributed applications face challenges with:

• Long-lived executions (minutes to days)
• Resource termination and process crashes
• Cross-process coordination
• State management across interruptions
• Partial failures and partial ordering

Traditional solutions like HTTP calls and job queues have limitations:

• HTTP requires maintaining connections
• Job queues lose execution context
• Both lack unified error handling
• Poor composability and developer experience

Key advantages of distributed async/await:

• Unified semantics for local and remote executions
• Built-in failure handling and recovery
• Resume semantics allow continuing from interruption points
• Maintains execution identity and integrity across processes
• Developer control over execution location and distribution

Core concepts:

• Durable promises with stable IDs for coordination
• Equivalent behavior between interrupted/resumed and continuous executions
• Composable functions and promises as universal abstractions
• First-class support for distribution in programming model

Practical benefits:

• Sequential-style code for distributed systems
• Automatic handling of partial failures
• No manual state management needed
• Natural expression of business logic
• Simplified development of long-running workflows