Deterministic Threshold Queries of Distributed Data Structures

Convergent replicated data types, or CvRDTs, are latticebased data structures for enforcing the eventual consistency of replicated objects in a distributed system. Although CvRDTs are provably eventually consistent, queries of CvRDTs nevertheless allow inconsistent intermediate states of replicas to be observed; and although in practice, many systems allow a mix of eventually consistent and strongly consistent queries, CvRDTs only support the former. Taking inspiration from our previous work on LVars for deterministic parallel programming, we show how to extend CvRDTs to support deterministic, strongly consistent queries using a mechanism called threshold queries. The threshold query technique generalizes to any lattice, and hence any CvRDT, and allows deterministic observations to be made of replicated objects before the replicas’ states have converged.