IFCA: Index-Free Community-Aware Reachability Processing Over Large Dynamic Graphs

Reachability is a fundamental graph operator. State-of-the-art index-based reachability processing frameworks can efficiently handle static graphs, but the recent advent of dynamic graph data poses new challenges. To address these challenges, we propose an index-free, community-aware (IFCA) reachability processing framework inspired by efficient Personalized PageRank approximation algorithms, which identifies community structures on-the-fly to accelerate query processing. On top of it, we devise a community contraction technique to bridge the gap between vertices in distinct communities, and a cost-based strategy selection procedure to efficiently handle the resulting reduced graph. We conduct experiments with realistic query workloads over large-scale real dynamic graphs, showing our approach’s superior efficiency compared with index-based and index-free state-of-the-art methods.