Java code using redis as a graph adjacency list. 
For querying neighbours, got response times near a milli second.
Neo4j is much faster, probably because the DB was embedded in the same jvm, no network overhead as in the case of Redis.