State machine replication (SMR) is a fundamental technique for implementing stateful dependable systems. A key limitation of this technique is that the performance of a service does not scale with the number of replicas hosting it. Some works have shown that such scalability can be achieved by partitioning the state of the service into shards. The few SMR-based systems that support dynamic partitioning implement ad-hoc state transfer protocols and perform scaling operations as background tasks to minimize the performance degradation during reconfigurations. In this work we go one step further and propose a modular partition transfer protocol for creating and destroying such partitions at runtime, thus providing fast elasticity for crash and Byzantine fault tolerant replicated state machines and making them more suitable for cloud systems.

Abstract. Despite the fact that cloud computing offers a high degree of dynamism on resource provisioning, there is a general lack of support for managing dynamic adaptations of replicated services in the cloud, and, even when such support exists, it is focused mainly on elasticity by means of horizontal scalability. We analyse the benefits a replicated service may obtain from dynamic adaptations in the cloud and the requirements on the replication system. For example, adaptation can be done to increase and decrease the capacity of a service, move service replicas closer to their clients, obtain diversity in the replication (for resilience), recover compromised replicas, or rejuvenate ageing replicas. We introduce FITCH, a novel infrastructure to support dynamic adaptation of replicated services in cloud environments. Two prototype services validate this architecture: a crash fault-tolerant Web service and a Byzantine fault-tolerant key-value store based on state machine replication.

No abstract

No abstract