Zeno: eventually consistent Byzantine-fault tolerance

摘要: Many distributed services are hosted at large, shared, geographically diverse data centers, and they use replication to achieve high availability despite the unreachability of an entire center. Recent events show that non-crash faults occur in these may lead long outages. While Byzantine-Fault Tolerance (BFT) could be used withstand faults, current BFT protocols can become unavailable if a small fraction their replicas unreachable. This is because existing favor strong safety guarantees (consistency) over liveness (availability). This paper presents novel state machine protocol called Zeno trades consistency for higher availability. In particular, replaces (linearizability) with weaker guarantee (eventual consistency): clients temporarily miss each other's updates but when network stable states from individual partitions merged by having agree on total order all requests. We have built prototype our evaluation using micro-benchmarks shows provides better than traditional protocols.