CODEWITHSUNDEEP
neo4jcypher
Jack Daniel
Jack Daniel

Reputation: 2611

Cypher - Order By Degrading the Query Performance

I am working on a graph DB where an organisation contains 1 Billion Employee records. If I try to order the employees on their performance and get the top 1000 efficient employees, the query is not returning the results. How can I fine tune this.

My query: 

    MATCH (org:Organization{org_guid:"12345"})-[r:EMPLOYED_BY]-(emp:Employee)
    WITH org,r,emp ORDER BY coalesce(r.efficiency,0) DESC
    LIMIT (1000)
    RETURN emp

Upvotes: 1

Views: 271

Answers (1)

William Lyon
William Lyon

Reputation: 8546

Currently in Neo4j ODRER BY is not able to use an index, which means it will require a scan, which can result in poor performance for large datasets.

There is an APOC procedure, however, that is able to use an index for range search and maintaining order. You would need to refactor your datamodel slightly as you cannot create an index on a relationship property. If instead you wanted to order by efficiency as a property on :Employee:

First, create an index on :Employee(efficiency):

CREATE INDEX ON :Employee(efficiency);

Then,

CALL apoc.index.orderedRange('Employee', 'efficiency', 0, 99999, false, 1000) 
YIELD node AS employee
RETURN employee;

Another slight complication is that you want descending order, so you would need to store the negative of the efficiency value to get the results in descending order.

Upvotes: 1

Related Questions