Project
AOEE: a social-graph cache
A TAO-inspired in-memory relationship cache for social-graph queries, with compressed posting lists and fast set operations for friend-of-friend queries and feeds.
AOEE (Attribute Object Enterprise Edition) is my take on the idea behind Facebook’s TAO: an in-memory cache of relationships that answers social-graph queries — followers, mutual friends, friend suggestions, feeds — in well under a millisecond.
What it does
-
Constant-time edge lookups and fast intersection and union for "mutual friends" queries.
-
Friend-of-friend (two-hop) queries with scoring for suggestions.
-
Sharding by consistent hashing for horizontal scale.
-
Compressed posting lists that pick an encoding automatically: a plain vector under 128 entries, delta-varint up to about 4,000, bit-packed blocks beyond that, and Roaring bitmaps for huge or dense lists.
How it is built
The server is Rust, serving gRPC; a Java client library and a Spring Boot proxy provide a REST API and dataset loading; and a React UI offers a graph explorer and edge manager. Each posting list combines an unsorted write buffer with compacted, sorted, encoded segments.
AOEE is the graph layer behind my enterprise social platform.