Consistent Updates for Scalable Microservices
summary
Presents a formal framework and the first algorithms that guarantee atomic-looking live updates for rolling microservice upgrades by routing requests based on semantic properties; builds an update manager that inspects requests and uses commutativity and forward/backward compatibility to rewrite mixed-mode traces into client-equivalent atomic ones, proves correctness with a partial-order/consistent-cut inductive proof and ranking measure, cites that 63% of real failures stem from cross-version interactions, and proves an impossibility result ruling out oblivious mixed-mode consistency.
AI-written summary · about
chapters (15)
- 0:00Consistent updates intro
- 1:05Scalable microservices setup
- 3:16Mixed mode dangers
- 5:46Update consistency framework
- 7:56Update consistency defined
- 9:19System behavior model
- 11:09Client equivalence in updates
- 12:35Commutativity update algorithms
- 14:18Partial order proofs
- 16:05Inductive proof method
- 18:54Oblivious algorithms impossible
- 20:20Summary and future research
- 21:03Q&A availability
- 23:10Q&A client groups
- 24:13Q&A worker workload
related talks
35mConsistent Updates in a Distributed System without Transactions
39mManaging Consistency, State and Identity in Distributed Microservices
27mData Consistency in Microservices Architecture
35mBuilding Scalable Stateful Services
51mResponsible Microservices
32mEscaping Version Skew: Formalizing Compatibility in a World of Partial Rollouts
48mCo-ordinating data changes in distributed systems
24mPatterns of Evolution from Monolith to Microservices