Is there a speed difference between versions?

[J_J_Dekker]

Hi there,

I have been building a model in the latest neo4j version 4.2.3. Using a simple model to upload a matrix for 13 x 30000 values as row -> column value. Querying this graph was fast in terms of 2-4ms for all values on my laptop. However I wanted to connect this model to a larger existing graph. This graph was made in version 3.5.24. After connecting some nodes and performing the same query as before (totally ignoring the new part of the graph), queries took many times longer. Instead of 2-4ms it was 10-30 minutes for a query.

Does this have to do with the model? or is the neo4j database version to blame?

Kind regards,
Julian

[dana_canzano]

if both the 3.5.24 and 4.2.3 environments were identical I would not expect that 4.2.3 had a 100x + improvement. Something else must be in play here to cause such a wide performance difference.

1. do both environments have the same total RAM?
2. are both configured with the same dbms.memory.pagecache.size, dbms.memory.heap.init_size and dbms.memory.heap.max_size
3. if you preface the query with profile do you get the same output, i.e. is the query plan the same?
4. do both environments have the same number and design for indexes? i.e. run call db.indexes();
5. do both environments have the same graph size?

[J_J_Dekker]

Hi there,

Thanks for the response. I just started remaking my model so it might take some time before I can test db.indexes and profile. I will get back to you with that in a bit. I do know that the 3.5 graph uses lucene indexes which I think are older.

The graph size for the older graph is bigger in total but the part I'm querying should be smaller than the 4.2 graph.
Heap init_size and heap_max_size and total RAM is the same.

Kind regards,
Julian