SimpleChubby: a simple distributed lock service

We implement a distributed lock service called SimpleChubby similar to the original Google Chubby lock service[1]. It intends to help implement distributed synchronization. SimpleChubby exposes a simple Unix-like file system interface combined with lock primitives and event subscription. Leveraging such interface, we build three distributed synchronization applications, namely leader election, group membership, and distributed barriers. SimpleChubby uses a server-client architecture, and we put great efforts to achieve fault tolerance where [1] does not provide many details. Server fault tolerance is achieved by combining the Viewstamp protocol [4] which we implement from scratch and a server-client cooperation mechanism to recover non-persistent server states. Server failure is transparent to users, and client failure is handled correctly by reclaiming its holding locks. In the following, SimpleChubby API and its semantics are described in section 2, the system architecture and normal-case processing mechanism are described in section 3, the fault tolerance mechanism is discussed in section 4, and in section 5, three applications are shown to illustrate how SimpleChubby API is used to build distributed synchronization applications, and to give a detailed description about the system workflow.