Doing More with Less: Orchestrating Serverless Applications without an Orchestrator
summary
Unum implements orchestration as a library that compiles Step Functions graphs into an IR and runs inside user Lambdas using DynamoDB (and Firebase on Google Cloud), decentralizing control so each function executes local orchestration instructions, uses a set-based fan-in to reduce reads, and checkpoints results for end-to-end exactly-once semantics. Evaluation vs AWS Step Functions shows Unum delivers up to 2x faster latency and is about 9x cheaper for representative applications, with most Unum cost coming from the datastore.
AI-written summary · about
chapters (9)
- 0:00Introduction to serverless
- 1:07Challenges in orchestration
- 3:04Drawbacks of current methods
- 4:39The case for Unum
- 6:23Design and implementation
- 10:00State and consistency
- 11:45Execution guarantees
- 13:44Performance evaluation
- 17:23Conclusion
related talks
51mChoreography vs Orchestration in serverless microservices
46mChoreography vs Orchestration in serverless microservices
26mOrchestrating Data Workflows Using a Fully Serverless Architecture
23mOrchestration Lightning: Rapid-Fire AWS Step Functions Patterns
29mOrchestrating complex (not complicated) tasks using AWS serverless and Python
21mServerless - Centralizing Data and Orchestrating Events Without Servers
59mFunction Orchestration in the World Of Serverless
38mChoreography vs Orchestration in serverless microservices