Help with Hardware Sizing for Neo4j Solution

Hello i need help and recommendations about Hardware that i need to create a solution running Neo4j as Database.

Actually I have 2 nodes and 1 relationship between this nodes.

  1. Node Clients (10,000,000) with 10 properties in each node.
  2. Node Products (90,000,000) with 10 properties in each node.
  3. RelationShip (75,000,000) with 10 properties in each relationship.

I want a calculation of a hardware that can store
2 , 000,000,000 of nodes with 10 properties in each node
3,000,000,000 of relationships with 10 properties in each node

Actually i dont need an Real-time response, but also it can be a possible requeriment.
can someone helpme giving me a reccomendation of hardware to have this information with neo4j with a real time response and without response in real time?

1 Node plus 10 properties will cost you about 140 bytes.
So 2 Billion of them is 280 gigs on disk.
1 Relationship plus 10 properties will cost you about 160 bytes.
So 3 Billion of them is 480 gigs on disk. Add them up and you have 760GB. Plus indexes plus additional internal data structures and you are looking closer to 850GB.

Now we typically recommend 1.2 to 1.5 times the amount of RAM that you have on disk FOR THE WORKING SET. Which is important. If only 10% of the graph gets queries on a particular day, then you would size for 85GB instead of 850GB.

Also, it's rare for relationships to have so many properties, so these numbers may be higher than you really need.

1 Like