Download
FlockDB is a specialized graph / adjacency-list storage system designed for high performance in large-scale, low-latency, real-time environments. It was developed at Twitter to store social graph data (followers, following, blocks, etc.) and secondary indexes. FlockDB emphasizes horizontal scalability, replication, and support for high rates of writes and updates, as well as efficient paging through very large result sets. It is not a general graph database in the sense of supporting complex multi-hop traversal queries or sophisticated graph algorithms; instead it focuses on the core problem of storing and querying directed edges with attributes such as sort order, state (normal, archived, removed), and position. Edges are stored both in forward and backward directions to facilitate queries in both directions. The project is now archived and in read-only mode, meaning it's no longer actively maintained by Twitter.
Features
- Supports massive adjacency lists for directed graphs (source → destination and destination → source stored)
- Supports add / update / remove of edges at high rate (writes) and paging through large result sets (reads)
- Ability to mark edges as “archived” (cold storage) and restore them later
- Sort position on edges to allow ordering (for example by timestamp)
- Horizontal scaling and replication for fault tolerance and performance
- Designed to be simpler than full graph databases: no multi-hop / graph walking, minimal feature scope, optimized for adjacency queries
Project Samples
