I see that the best recommended AWS instance type for Neo4J is usually t2 or t3 and — sometimes — m2 and m3.
However, as I'm having problems mainly with memory consumption I'm considering getting an R6g instance, which is memory-optimized (and has less CPUs) — https://aws.amazon.com/ec2/instance-types/
What do you think about this? Which instance would be the best?
I think the right answer is "it depends".
can you tell us about the db, the instance it is on, the configuration you are using and the memory consumption issue?
how are you factoring in the data aspects specific to your graph db? (nodes/rels, connectivity, on disk size, cache requirements, etc)
what are the primary usage patterns? (number of users, types of queries...)
Hello, I need some advice regarding this as well.
We expect to start with 200 million relationships (around 5 properties) and 50 million nodes (around 5 properties as well).
We won't have many users for now, but we have writes to the db every hour, which will be around 10k relationships.
Our use case is around account balances, so the heaviest query will be getting the balances of all accounts.
The rest will mostly be around querying data by timeframe, or node. Which shouldn't be more than 10k relationships.
Is it preferred to try and keep it all in memory for now?
Since the growth rate is a lot slower than the initial graph size?
Happy to answer more questions if need.