Ask HN: Graph database design guidelines

2 points

13 years ago

What would be various design guidelines for property graph databases like Neo4j? In most examples I have come across, nouns/entities are stored as nodes and edges/relationships are generally verbs (in some form). Is this typical?

Which information should be stored as properties and which as separate nodes? For example, consider a red shirt. Should this be stored as property Color=Red in node (Shirt), or as (Shirt)--hasColor-->(Red)?

Are there any recommendations for naming conventions? There may be tradeoffs between long property names and labels (self-documenting) and performance (more memory, longer string comparison operations)?

What other considerations are there that should not be overlooked?

Please share your experiences. What data models do you use for graph databases in your applications?

I could not find much information on the subject on the Internet other than [1] and [2] which do not say as much as I would like.

[1] http://stackoverflow.com/questions/6482741/graph-database-design-principles-general-principles-and-granularity-issue

[2] http://stackoverflow.com/questions/10753331/graph-database-design-methodologies