Intro To Graph Databases - Oxana Goriuc
Download as PPTX, PDF0 likes165 views
The document provides an introduction to graph databases and their applications, particularly through the Dgraph platform used by the payment security company Fraugster. It covers the basics of graph theory, the nature of graph databases, and specific operational details for deploying and querying Dgraph. Additionally, it discusses applications like fraud detection, social networks, and neural networks, along with practical tips for data loading and metrics.
Intro To Graph Databases - Oxana Goriuc
- 1. INTRO TO GRAPH DATABASES by Oxana Goriuc
- 2. www.fraugster.com About Fraugster: ● We are a payment security company fighting eCommerce payment fraud ● We are a team of 61 employees, of 26 nationalities ● The company was founded in 2014, based in Berlin, Rocket Tower
- 3. www.fraugster.com Agenda ● Graph Theory basics ● Graph Databases and their applications ● Dgraph Intro ● Dgraph Deployment ● Data Loading ● Query Language ● Visualization tool
- 4. www.fraugster.com NodeEdge Entity = Connected component What is a graph? Graph is a collection of nodes and relationships between them txn3 txn4 txn5 txn6 txn2 txn1 phone email credit card
- 7. www.fraugster.com What is a graph database? ● Graph Database uses graph structures to store and represent the data ● NoSQL ● Support ACID ● Edges (relationships) have type and direction and can have properties ● Connected component can be usually retrieved with one query Social Network Graph
- 8. www.fraugster.com Graph Applications ● Computer Science (flow of computation) ● Map services (intersection of streets is considered to be a node) ● Social Network (Facebook, Twitter) ● Fraud detection (eCommerce, financial fraud) ● Recommendation systems (in eCommerce) ● Search engines (Knowledge Graph) ● Neural networks https://blog.finxter.com/graph-applications/ https://blog.finxter.com/graph-applications/
- 9. www.fraugster.com DGraph Intro ● Written in Go ● Fast and scalable ● Distributed ● Open source ● Build-in visualization tool ● Supports high availability ● Flexible schema ● Facebook Graph Query Language
- 10. www.fraugster.com DGraph Deployment (on Linux) 1. Install docker and docker-compose 2. Save the snapshot into docker-compose.yml 3. Run “docker-compose up -d” 4. Done! https://docs.dgraph.io/get-started/
- 11. www.fraugster.com Data loading Loading into DGraph: Bulk Loader / Live Loader RDF TXN DGraph
- 12. www.fraugster.com Data loading tips Instead of finding all the existing links in the data before loading (which can be done by merging the table with itself multiple times or loop through the data), try to represent your linking elements as separate nodes and make use of deduplication, which happens in Dgraph automatically
- 13. www.fraugster.com Data loading tech metrics ● Donec risus dolor porta venenatis ● Pharetra luctus felis ● Proin vel tellus in felis volutpat Max memory usage RDF row count/size 51.8M, 3.8 GB 225.9M, 15.8 GB Dataset 11.9 millions2.9 millions of txns Loading time 7m45s RSS: 6.5 GB VMS: 25.2 GB 40m47s RSS: 8.6 GB VMS: 59.1 GB
- 14. www.fraugster.com DGraph Query Language How to query Dgraph: ● Graphical UI ● Client (Go, Python, Java …) { bladerunner(func: eq(name, "Blade Runner")) { uid name initial_release_date netflix_id } } Example: retrieve the information about the movie “Blade Runner” Output: ● JSON ● Graph Visualization ● Every level should be enclosed with curly brackets ● A function always starts with a name ● uid means unique identifier of a node in Dgraph ● Specific to graph DB queries: recurse, shortest path ● Output of a function can be saved in a variable