Remove-Win: a Design Framework for Conflict-free Replicated Data Collections


Abstract in English

Internet-scale distributed systems often replicate data within and across data centers to provide low latency and high availability despite node and network failures. Replicas are required to accept updates without coordination with each other, and the updates are then propagated asynchronously. This brings the issue of conflict resolution among concurrent updates, which is often challenging and error-prone. The Conflict-free Replicated Data Type (CRDT) framework provides a principled approach to address this challenge. This work focuses on a special type of CRDT, namely the Conflict-free Replicated Data Collection (CRDC), e.g. list and queue. The CRDC can have complex and compound data items, which are organized in structures of rich semantics. Complex CRDCs can greatly ease the development of upper-layer applications, but also makes the conflict resolution notoriously difficult. This explains why existing CRDC designs are tricky, and hard to be generalized to other data types. A design framework is in great need to guide the systematic design of new CRDCs. To address the challenges above, we propose the Remove-Win Design Framework. The remove-win strategy for conflict resolution is simple but powerful. The remove operation just wipes out the data item, no matter how complex the value is. The user of the CRDC only needs to specify conflict resolution for non-remove operations. This resolution is destructed to three basic cases and are left as open terms in the CRDC design skeleton. Stubs containing user-specified conflict resolution logics are plugged into the skeleton to obtain concrete CRDC designs. We demonstrate the effectiveness of our design framework via a case study of designing a conflict-free replicated priority queue. Performance measurements also show the efficiency of the design derived from our design framework.

Download