Logic and Algebras for Distributed Computing: Applications and Open Questions
summary
Vision and prototypes present a concrete stack: translate Daedalus programs into a Hydro/Hydroflow IR, compile with e-graph style rewrites (Egg/Egglog) into per-node Rust implementations, and deploy with automated partitioning and replication. Key claims: free replication via semilattice CRDTs yields massive gains (Anna shows ~700x throughput under contention), Hydroflow matches handwritten C++ latency/throughput, and automated transformations reproduce or exceed hand-tuned Paxos performance; several pieces remain research questions (monotonicity, termination, cost-based search).
AI-written summary · about
chapters (9)
- 0:00Logic and algebras for cloud computing
- 1:07Cloud lacks canonical model
- 8:51Free replication and CALM theorem
- 21:54Semilattices and CRDTs
- 30:24CALM theorem and coordination-free
- 41:26Logic and algebra for distributed computing
- 47:04Automatic partitioning of programs
- 53:35Open questions in distributed algebras
- 1:00:13When clocks are needed
related talks
46mCoordination-Free Computations
52mI See What You Mean
41mProgramming Distributed Systems
17mBring Your Own Data Structures to Datalog
52mData Over Code: A Short Hyperbole
52mThe Computer Science behind a modern distributed data store
21mTimely Dataflow
28mThe Art of the State: Serverless Computing and Distributed Data