GLogS: Interactive Graph Pattern Matching Query At Large Scale
summary
GLogS is a distributed system that compiles declarative graph pattern queries (supports Cypher-like languages) into dataflow programs, builds a GLog catalog of pattern frequencies, and chooses hybrid execution plans using those higher-order statistics. It reports 51× faster vs Neo4j on a single machine and avoids TigerGraph’s manual tuning and 1–3 minute native compilation by doing plan generation online. A stratified specification shrinks graphs for frequency estimation, and the deployed system scales across 16 back-end nodes with near-linear speedups.
AI-written summary · about
chapters (14)
- 0:00GLogS introduction
- 3:56Graph DB limitations
- 6:45GLogS architecture
- 10:14GPM optimization
- 13:22Pattern frequency estimation
- 16:57G-Log operations
- 17:40Plan generation
- 19:03Experimental setup
- 19:49Performance vs Neo4j/TigerGraph
- 20:31Higher-order statistics
- 21:12Stratified graph impact
- 21:35Scalability results
- 22:17Performance trends
- 23:01Summary
related talks
56mGraph Analytic Systems
06mUnlocking graph analytics in DuckDB with SQL/PGQ
30mBeyond the Silos Achieving Data Harmony With Graph Based ER
39mThorny path to the Large Scale Graph Processing
24mLarge Scale Graph Computation with Apache Giraph
39mThorny path to the Large Scale Graph Processing
30mHow Graph Data Science Can Turbocharge Your Knowledge Graph
48mAnalytics on Large Graphs in S3: Tools, Approaches and Formats for OLTP and OLAP