Storing your Graphs
Olivier Yiptong : oyiptong@mozilla.com
Triple Stores
- Store triples, i.e. relationships
Bob :LIKES Alice
Alice :LIKES Charlie
Charlie :IS NotSingle
- SPARQL standardized
- Mostly OLAP. AllegroGraph claimed OLTP-capable
- A specific kind of graph
Graph Databases
- Index-free adjacency
- No standard storage strategy
- Various models
- Most seem focused on OLTP
Other
Roll your own on top of:
image: http://eng.orcadigital.com/choosing-a-database-for-the-cloud/
A look at Graph Databases
http://blog.octo.com/en/graph-databases-an-overview/
No standard implementation
http://www.slideshare.net/thobe/an-overview-of-neo4j-internals
No standard implementation
http://www.slideshare.net/sgomezvillamor/dex-presentation-for-gdm-2011-7649
Implementation matters!
No Free Lunch
Massive Scale
- Massive tradeoffs
- Pregel, Giraph, Hama etc for offline processing
- Titan? Not exactly transactional processing
- Data rarely massive
- Massive scale == Massive pain
Which Graph DB then?
It depends!
Perhaps none!
- Facebook employees on Quora answer
What exactly is Unicorn?
Cache-upon-cache-upon-cache-upon-cache
"The Truth" stored in MySQL
- LinkedIn: same story
The Big Data Ecosystem at LinkedIn
Polyglot Persistence
GraphDB's can be very useful
- Usage scenarios
- Your data
- Types of computation
- Great tool in your arsenal
Thanks!
Olivier Yiptong : oyiptong@mozilla.com
Graph Databases
By Olivier Yiptong