You're right about caching. I wasn't sure if graph databases took the typical NoSQL approach of massive non-normalized tables to avoid computational overhead.
SQL doesn't scan either -- tables are indexed via either a hash (O(1)) or a tree (O(log edges)). While I do realize the speed advantage of a specialized structure, I feel this could be better implemented as a specialized index/storage mechanism in SQL than as an entirely new database system.
Comments
You're right about caching. I wasn't sure if graph databases took the typical NoSQL approach of massive non-normalized tables to avoid computational overhead.
SQL doesn't scan either -- tables are indexed via either a hash (O(1)) or a tree (O(log edges)). While I do realize the speed advantage of a specialized structure, I feel this could be better implemented as a specialized index/storage mechanism in SQL than as an entirely new database system.