neo4j is a graph database. this is based off of graph theory and handles every object as nodes and edges. this was kind of pioneered by the sparql movement awhile back, but wikidata uses graph databases as their backend for many things.
Post by twetch#7453
though neo is quite interesting, i personally have never seen it scale it while outperforming relational databases. there is decades of knowledge for how to scale relational databases, the same can't be said for non relational (excluding columnstore/kvp)
twetch#7453 6.6y
a graph db is merely abstraction
What the chain says
- Block
- 660 013
- Time
- 2020-11-05T19:57:45Z
- Signer
- 1F8LZF7xb8a7aSnGV226bXECALFP9u8gsQ
- App
- twetch
- Type
- post
- Content type
- text/plain
- Name in tx
- twetch#7453
Fields the transaction did not carry are omitted. Open the payload to see the bytes as stored.
Signed by
1F8LZF7xb8a7aSnGV226bXECALFP9u8gsQ VerifiedReposted by (1)
Replies (11)
document store databases allow you to "basically store anything". this freedom makes it extremely difficult to analyze the results, especially with large amounts of data. graph databases handle wildly recursive relationships, but that often isn't good
if you really care about graph like data, geospacial indices within relational systems are probably one of the most interesting methods you could use to scale
twetch#7453 6.5y
i would love to see someone pull off a graph db equivalent using spacial indices in SQL. really wouldn't be that hard, could probs beat neo4j performance on high d recursion pretty easily
i havent really looked into urbit's graph store yet, but i just don't think mongo is that interesting or useful for scalable systems. i hope that there is more to it than that tbqh
I use foundationdb also. It’s not even meant for other than a datacenter really, running it on localhost isn’t it.
the reason this is important is bc mongo isn't particularly difficult to implement and i wouldn't call it a graph. frankly, it's just not that useful for scalable applications imo. i have not looked at urbit's yet implementation though so i can't judge
twetch#7453 5.9y
interesting that it's more like mongo than neo
To a noob like me this kinda sounds like #21e8
hahahaha in many ways you're spot on. indices off of a hash based placement model are basically that, just in abstract space https://infolab.usc.edu/DocsDemos/yao_vldbjournal.pdf (i *think h/t @1623)