Rethinking Safe Consistency in Distributed Object-Oriented Programming
summary
ConSysT is an extension to Java that models a replicated store with Ref references annotated by consistency (examples: @Ev for eventual) and supports sequential, causal, and eventual levels. A static type system forbids unsafe flows from weak to strong consistency so strong-consistent operations never depend on stale weak data; the paper proves type safety and non-interference. Implementation and middleware run three configs (all-sequential, all-weak, mixed) across five case studies and show mixed configs linearly reduce latency relative to all-sequential.
AI-written summary · about
chapters (11)
- 0:00Start
- 2:20Consistency?
- 3:10Which consistency level to choose for my application?
- 5:14Object-Oriented Design
- 7:08Multiple Consistency Levels
- 7:41Consistency Leaks
- 10:20Mixing of Consistency Levels class Concert
- 11:14Formalization
- 11:54Evaluation
- 12:45Case Studies
- 14:00Effect of Strong Consistency - Message Groups
related talks
14mBolt-On Strong Consistency: Specification, Implementation, and Verification
17mMixed & Verified Consistency with Propel and ConOpY
25mThe shell game called eventual consistency
41mProgramming Distributed Systems
41mTowards language support for distributed systems
24mArbitration-Free Consistency Is Available (and Vice Versa)
52mEventual Consistency in a Stateful Service
22mHow to Enforce and Verify Invariants in Weakly Consistent Databases