ClickHouse Test: v2

Dataset

Use the OnTime Dataset (2020–2022 data only) to ensure each Neo4j query finishes within 1 hour.

Information of the dataset (2020–2022 data only):

• 12282219 rows in total.
• 435 nodes, 12282219 edges.
• csv files: about 5GB.
• In ClickHouse: 925.28 MiB

Machine Configuration

ClickHouse Server & PuppyGraph with No Cache

An AWS EC2 instance of type m6a.2xlarge, with 8 vCPUs, 32GB RAM and 128GB storage (gp3), running Clickhouse version 25.6.2.5. Note that ClickHouse Server & PuppyGraph instance are in the same machine.

PuppyGraph with Full Cache

Two AWS EC2 instances of type m6a.2xlarge, with 8 vCPUs, 32GB RAM and 128GB storage (gp3), running PuppyGraph version 0.81.

Neo4j

An AWS EC2 instance of type m6a.2xlarge, with 8 vCPUs, 32GB RAM and 128GB storage (gp3), running Neo4j version 2025.06.2-community.

Results