Asynchronous Complete Garbage Collection for Graph Data Stores

Graph data stores are a popular choice for a number of applications: social networks, recommendation systems, authorization and control access, and more. Such data stores typically support both distribution and replication of vertexes across physical nodes. While distribution provides better load balancing of requests, replication is necessary to achieve improved availability and performance. However, most of these systems still manage replicated memory by hand, resulting in expensive efforts to fix dangling references and memory leaks to ensure referential integrity. In this paper, we present a novel Garbage Collection (GC) algorithm that safely eliminates both acyclic and cyclic garbage with support for replicated data. Our algorithm provides minimum impact on applications' performance, is completely decentralized and has no coordination requirements. We evaluate our approach against previous solutions and show that our solution is efficient while imposing very little overhead on applications' performance.