Memgraph write speed getting slower as data gets bigger about memgraph

[SimonCigoj]

Interesting ... we are experiencing occasional slow queries using MERGE where we create a relationship to parent node and also on MATCH where we update a property in a node .... it randomly takes over a second and this time increases linearly with the number of nodes in the graph. On the updated node we have 1 index

for example this query :

MATCH (ma:Parent {id:$id}) -[]- (m:Child {id:$id}) SET m.status = $status

occasionally takes a second and more and increases linearly with the number of nodes .... status does not have an index, Child id and Parent id both are indexed.

Till date I have no % of how many requests are slow but I would say around 5% at least. For us is really important to have consistent times

Sure it probably depends on the graph structure and the may not be the same as mentioned in this ticket. But the interesting part is that the insert time increases linearly as the graph keeps getting bigger.

Let's say that 90 percentile of the queries execute in 4 ms average but the 95 percentile is a second and more.

Currently we have around 1 million nodes in graph 2 million relationships 36 indexes and 10 constrains ... there are 14 different type of nodes/labels.

We are using Memgraph 2.13.0 and java 21 on Quarkus. for the driver we use

 <dependency>
      <groupId>io.quarkiverse.neo4j</groupId>
      <artifactId>quarkus-neo4j</artifactId>
      <version>3.4.1</version>
</dependency>

@antejavor

[antejavor]

Hi @SimonCigoj, Thanks a lot for providing extra context on the issue and explaining your scenario.

Do you have any other issues except this one with the unstable write performance? How is the Memgraph experience in general?