Neo4j for Solving Graph-Based Problems
summary
The talk explains how Neo4j was used to replace a two-hour exhaustive route enumeration with a production pipeline that delivers a median ~340 ms search time by importing CSVs, modeling days as nodes, and compressing a terabyte of precomputed routes into a 4 GB graph. It describes using two-way (bidirectional) traversal plus depth-first search with boundary pruning and dynamic programming to quickly find the top-200 routes, with imports completing in minutes, daily batches of ~1M unique directions finishing in about an hour on a larger node, and custom Java server-side extensions implementing the optimized two-sided search logic.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 0:42Company goals
- 3:31Old solution problems
- 4:59Why graph databases
- 6:00Graph database basics
- 8:11Storage optimization
- 9:32Neo4j features
- 11:58Data import to Neo4j
- 13:22Data modeling
- 15:51Bidirectional search
- 18:02Pathfinding algorithms
- 19:28System architecture
- 20:11Deployment
- 21:36Results and performance
- 22:38Scalability benefits
- 23:01Graph database recommendations
- 24:03Cypher examples
- 26:56Q&A
related talks
32mNever use a graph database!
55mLet Me Graph That For You
28mNeo4j for Beginners: Pros and Cons, Based on Social Networks
29mTraversing from Zero to Production with Graphs
21mSerious Network Analysis Using Hadoop & Neo4j
30mPerformant GraphRAG Improving Neo4j to Drive Autonomous AI ...
27mSpeed and Precision: Mastering Graph Traversal With Quantified Path Patterns
28mInterview with