Hampa: Solver-Aided Recency-Aware Replication

Abstract

Replication is a common technique to build reliable and scalable systems. Traditional strong consistency maintains the same total order of operations across replicas. This total order is the source of multiple desirable consistency properties: integrity, convergence and recency. However, maintaining the total order has proven to inhibit availability and performance. Weaker notions exhibit responsiveness and scalability; however, they forfeit the total order and hence its favorable properties. This project revives these properties with as little coordination as possible. It presents a tool called that given a sequential object with the declaration of its integrity and recency requirements, automatically synthesizes a correct-by-construction replicated object that simultaneously guarantees the three properties. It features a relational object specification language and a syntax-directed analysis that infers optimum staleness bounds. Further, it defines coordination-avoidance conditions and the operational semantics of replicated systems that provably guarantees the three properties. It characterizes the computational power and presents a protocol for recency-aware objects. uses automatic solvers statically and embeds them in the runtime to dynamically decide the validity of coordination-avoidance conditions. The experiments show that recency-aware objects reduce coordination and response time.

Synthesizing CRDTs from Sequential Data Types with Verified Lifting

“…Few previous systems have attempted to directly apply search-based program synthesis to the space of replication. Two that stand out are Hamsaz [Houshmand and Lesani 2019] and Hampa [Li et al 2020]. Hamsaz uses programmer-provided invariants to synthesize custom consistency protocols for the replication of shared data structures.…”

ECROs: building global scale systems from sequential code

“…To ensure state convergence, ECROs deterministically order concurrent noncommutative operations at all replicas. In contrast, existing approaches [Houshmand and Lesani 2019;Kaki et al 2018;Li et al 2012Li et al , 2018Li et al , 2020Sivaramakrishnan et al 2015] coordinate these operations unnecessarily.…”

“…However, the aforementioned RDTs do not consider application invariants. Hamsaz [Houshmand and Lesani 2019] and Hampa [Li et al 2020] statically analyze data types to derive coordination protocols that guarantee state convergence and preserve invariants. The derived protocols coordinate all non-commutative and unsafe method calls and are thus more conservative than ECROs.…”

“…However, those approaches do not aid with building correct RDTs which is known to be difficult and error-prone [Baquero et al 2017;Kaki et al 2018;Kleppmann and Beresford 2017]. Hamsaz [Houshmand and Lesani 2019] and Hampa [Li et al 2020] derive coordination protocols from a data type's specification. Overall, the aforementioned approaches tend to be conservative, leading to unnecessary coordination (cf.…”

“…Similarly, unsafe operations may run concurrently if a safe reordering exists. In contrast, existing approaches [Balegas et al 2015;Houshmand and Lesani 2019;Li et al 2012Li et al , 2018Li et al , 2020 impose coordination on all non-commutative operations and unsafe operations. Finally, ECROs ignore causality between unrelated commutative operations while prior works always execute operations in an order that is compatible with causal order.…”