Columnar storage and list-based processing for graph database management systems

We revisit column-oriented storage and query processing techniques in the context of contemporary graph database management systems (GDBMSs). Similar to RDBMSs, GDBMSs support read-heavy analytical workloads that however have fundamentally different data access patterns than traditional workloads. first derive a set desiderata for optimizing processors GDBMS based on their patterns. then present design columnar storage, compression, these desiderata. In addition showing direct integration existing from we also propose novel ones are optimized GDBMSs. These include list-based processor, which avoids expensive copies block-based under many-to-many joins, new structure call single-indexed edge property pages an accompanying ID scheme, application Jacobson's bit vector index compressing NULL values empty lists. integrated our into GraphflowDB in-memory GDBMS. Through extensive experiments, demonstrate scalability performance benefits techniques.