I'd welcome any insights and heuristics people might have regarding the 
following:

I'm looking to use neo4j for domain modeling, and some of the info I want 
to capture is clearly 'graph-centric' and will be used for specifying 
traversals, etc: I'll obviously be using core neo4j functionality for that, 
and there's no confusion or ambiguity there. But there will also be large 
chunks of data that from a graph perspective are  'just payload' and all I 
would be doing from neo4j's perspective is finding the data, reading it 
into the app, and making decisions based on the contents. 

The data would be messy and different in content and/or format for many 
different nodes, but we're probably talking dozens-to-one-hundred different 
key-value pairs with the values being ints or sentence-long strings. We're 
probably not talking more than 10 kilobytes of info per 'chunk of payload', 
so this isn't a matter of sheer volume, just messiness and lack of 
structure.

I was wondering which of the following options smells the best to the 
experts:
1. create a bunch of individual properties for each node as needed, and 
store it all in neo4j.
2. create a json-or-similar blob in a catch-all neo4j property for each 
node, store all this stuff in that one blob as needed read it out, and 
parse it in-app.
3. just store a UUID or key in neo4j, and use the key to store this messy 
data as a blob in some other noSQL store type. When I need the payload, I'd 
use neo4j to get the key, then grab the payload from the other store, then 
parse it in-app.

Any suggestions on which of the above is smart/fast/clean/scalable, etc? Am 
I missing a smarter alternative? And where there isn't enough info above to 
decide, what other criteria would help me make an informed decision?

Thanks,
  Tim.

-- 
You received this message because you are subscribed to the Google Groups 
"Neo4j" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to