An Approach For The Internet Of Things
summary
Argues for pushing computation and state to devices using CRDTs and a declarative lattice processing model (Erlang library) so disconnected sensors can compute, merge and track provenance deterministically. Describes delta state-based dissemination (Delta state-based CRDTs) with epidemic gossip (HyParView + Plumtree) to reduce bandwidth, targets scaling to 10,000 nodes, and shows real use cases: hospital-fridge alerts and an ad-impression counter (Rovio) where local increments merge correctly and servers perform the single exact-once actions.
AI-written summary · about
chapters (14)
- 0:00Start
- 1:00Edge Computation
- 4:38Ideal Execution
- 6:05Making Local Decisions
- 7:49Transitive Dissemination
- 11:31Consistency of Replicated Data
- 13:07Eventual Consistency
- 19:30Strong Eventual Consistency
- 23:49Lattice Processing
- 25:32Delta Base State Dissemination
- 29:52Broadcast Trees
- 32:50Data Flow Graph
- 35:04Divergence
- 37:41Challenges
related talks
41mDeclarative, Convergent Edge Computation
36mConflict-Free Replicated Datatypes (CRDT)
48mCRDTs: From sequential to concurrent executions
40mPractical Demystification of CRDT
34mUsing CRDTs Beyond Text Editors
21mCRDTs for Mortals
52mArchitectural Patterns for Distributed and Decentralized Data
39mEventually Consistent Data Structures