Run ClickHouse and measure performance

ClickHouse is a column-oriented database management system (DBMS) for online analytical processing of queries (OLAP).

You can use ClickBench to measure the processing time (query latency) of ClickHouse on Arm servers.

ClickBench is open-source software used to evaluate the performance of various database management systems for web analytics.

Before you begin

You will need an Arm server or an Arm based instance from a cloud service provider running a recent version of Ubuntu for Arm.

You will also need sufficient storage on the instance for the web-analytics dataset used for measuring ClickHouse performance, 500 GB is recommended.

Install ClickHouse

Install ClickHouse and start the server. For detailed installation instructions refer to the installation guide .

1. Install ClickHouse:

    

        
        
            curl https://clickhouse.com/ | sh
sudo DEBIAN_FRONTEND=noninteractive ./clickhouse install
        
    

2. Set the compression method for clickhouse-server to use zstd by running the commands:

    

        
        
            echo "
compression:
    case:
        method: zstd
" | sudo tee /etc/clickhouse-server/config.d/compression.yaml
        
    

3. Start the ClickHouse server:

    

        
        
            sudo clickhouse start
        
    

Run ClickBench

1. Clone the ClickBench repository:

    

        
        
            sudo apt install -y git wget curl
git clone https://github.com/ClickHouse/ClickBench.git
        
    

2. Navigate to the repository and create a table:

    

        
        
            cd ClickBench/clickhouse
clickhouse-client < create.sql
        
    

3. Load the benchmark data

The data file is very large and takes more than 10 minutes to download and uncompress.

    

        
        
            wget --continue 'https://datasets.clickhouse.com/hits_compatible/hits.tsv.gz'
gzip -d hits.tsv.gz
        
    

4. Import the data using clickhouse-client:

Importing the data takes more than 5 minutes.

    

        
        
            clickhouse-client --time --query "INSERT INTO hits FORMAT TSV" < hits.tsv
        
    

The data used is the Anonymized Web Analytics dataset .

5. Execute the benchmark

The script loops through each query three times. A total of 43 queries are run.

    

        
        
            ./run.sh
        
    

ClickBench results

When you execute the run.sh script, the query processing time for each individual query is displayed on the console.

The three comma separated values represent the query latency time for each of the three times the query is run.

    

        
        [0.002, 0.001, 0.001],
[0.028, 0.023, 0.022],
[0.066, 0.052, 0.052],

        
    

The summarized results are also saved to the results.csv file in the current directory. The results.csv file has 129 lines (43 queries each run 3 times).

You can try different types of hardware and compare results. For example, if you use AWS try c6g.2xlarge and compare to c7g.2xlarge to see the difference between AWS Graviton2 and Graviton3 processors.