This message was deleted.

great topic, Sandeep! I think it depends on your query pattern. The recommended way to decide between modeling methods is to list all query patterns your system will do, and see their costs in the GraphDB. i.e.

to find everyone who likes a particular store

• in intermediate node approach will be:

start from store vid, expand via at_store reversely, expand via likes reversely

• in another approach will be:

start from store vid, expand via sells, expand via likes reversely

The cost of the two are basically the same.

find which store's apples Angela likes

, the 2nd approach is cheaper as only one-hop traversal is needed, especially in NebulaGraph

find how many groceries that Angela likes at Walmart

, ditto So I would suggest listing all query patterns your system cares first. :)

Oh Okay! Thanks @wey :)