Is it OK to have millions of directed relationships between two nodes in Neo4j? Will it add to latency in fetching the data?

user_16478442 0
Viewed 117

We have millions of users of can make millions of transactions between them.

Lets say there is a user_A who pays money to user_B on the daily basis. We have a relationship SEND_MONEY_TO between A and B nodes.

A ---SENDS_MONEY_TO---> B

What will be the better design to accommodate this data in Neo4j.

Option A: We will create a new relationship edge every time a transaction happens.

Option B: We will keep a list of transactions as property of a same relationship edge and append the transaction details to existing list whenever a new transaction happens.

Our queries will look like:

a.) Find number of transaction between user_A and user_B in month of April 2021 where HDFC credit card is used.

b.) Find the total amount of transactions which involves user_A

We are open to any new approach as well.

graph-databases neo4j
2 Answers

I assume in the future, you may want to add other data to transactions, like the :Card :Device or :Platform on which it was triggered/Executed. As you probably know, you cannot create an edge between an edge and a vertex.

I would therefore recommend to use (:Transaction) vertices, with edges to the (:Account) vertices. You could use [:SENDER] and [:RECEIVER] types for the edges.

user_17349960

Here is how I've modeled this in the past. You might need to make a few adjustments to model it in Neo4j, but the general gist is there. Sample projects attached.

Because a customer might have multiple accounts

Customer -(CUSTOMER_ACCOUNT)- Account

Capture the directionality of the transactions

Account - (SEND_TRANSACTION)-> Transaction

Account <-(RECEIVE_TRANSACTION)-Transaction

Here I capture a running total of min_Send, min_receive, max_send, max receive, transaction_send_count, transaction_recive_count, etc

Account -(SEND_TO)-> Account

I've open-sourced the project here:

Sample Google Colab:

user_126845280
Related