Scalable coordination of a tightly-coupled service in the wide area

Abstract

Today's large-scale services generally exploit looselycoupled architectures that restrict functionality requiring tight cooperation (e.g., leader election, synchronization, and reconfiguration) to a small subset of nodes. In contrast, this work presents a way to scalably deploy tightlycoupled distributed systems that require significant coordination among a large number of nodes in the wide area. Our design relies on a new group membership abstraction, circuit breaker, to preserve efficient pairwise communication despite transient link failures.We characterize our abstraction in the context of a distributed rate limiting (DRL) service's deployment on a global testbed infrastructure. Unlike most distributed services, DRL can safely operate in separate partitions simultaneously, but it requires timely snapshots of global state within each. Our DRL deployment leverages the circuit breaker abstraction along with a robust gossipbased communication protocol to meet its demanding communication requirements. Through local and widearea experiments, we illustrate that DRL remains accurate and fair in the face of a variety of failure scenarios.

When Raft Meets SDN

“…Robust message exchange in SDN is fundamental for controller availability. Webb et al [29] propose a way of deploying tightly-coupled distributed system in wide area in a scalable way. It preserves efficient pairwise communication through an overlay network with gossip-based communication protocol.…”