3 ms·
Right, but unless you know the queries in advance there is always the risk of pathological queries that thrash the network. Naive example: in the movie dataset
by nl 10y ago
Right, but unless you know the queries in advance there is always the risk of pathological queries that thrash the network.
Naive example: in the movie dataset, if you partition by node type and have actors on one server and films on another a query like "find me all films with actors names starting with M who also starred in films with actors starting with N" will perform horribly, but if you partition by actor and film name it will be OK.
Titan (and I think most distributed Graph DBs) use pluggable distribution strategies and default to random to try to combat this problem.