Exasol vs ClickHouse Performance Comparison on TPC-H

Overview

The Baseline: Exasol

Exasol achieved a median query runtime of 67.9ms across all TPC-H queries, establishing a competitive performance baseline for analytical workload processing.

Clickhouse 25.9.3.48 - System Under Test

Installation & Configuration

# Stop existing RAID array at /dev/md0 if present
sudo mdadm --stop /dev/md0 2>/dev/null || true
# Clear RAID superblock on /dev/nvme1n1
sudo mdadm --zero-superblock /dev/nvme1n1 2>/dev/null || true
# Clear RAID superblock on /dev/nvme2n1
sudo mdadm --zero-superblock /dev/nvme2n1 2>/dev/null || true
# Create RAID0 array from 2 devices
yes | sudo mdadm --create /dev/md0 --level=0 --raid-devices=2 /dev/nvme1n1 /dev/nvme2n1
# Wait for RAID array /dev/md0 to be ready
sudo mdadm --wait /dev/md0 2>/dev/null || true
# Create mdadm configuration directory
sudo mkdir -p /etc/mdadm
# Save RAID configuration
sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf
# Format /dev/md0 with ext4 filesystem
sudo mkfs.ext4 -F /dev/md0
# Create mount point /data
sudo mkdir -p /data
# Mount /dev/md0 to /data
sudo mount /dev/md0 /data
# Set ownership of /data to ubuntu:ubuntu
sudo chown -R ubuntu:ubuntu /data
# Create ClickHouse data directory under /data
sudo mkdir -p /data/clickhouse
# Set ownership of /data/clickhouse to clickhouse:clickhouse
sudo chown -R clickhouse:clickhouse /data/clickhouse

# Update package lists
sudo apt-get update
# Install prerequisite packages for secure repository access
sudo apt-get install -y apt-transport-https ca-certificates curl gnupg

# Add ClickHouse GPG key to system keyring
curl -fsSL 'https://packages.clickhouse.com/rpm/lts/repodata/repomd.xml.key' | sudo gpg --dearmor -o /usr/share/keyrings/clickhouse-keyring.gpg
# Add ClickHouse official repository to APT sources
ARCH=$(dpkg --print-architecture) && echo "deb [signed-by=/usr/share/keyrings/clickhouse-keyring.gpg arch=${ARCH}] https://packages.clickhouse.com/deb stable main" | sudo tee /etc/apt/sources.list.d/clickhouse.list
# Update package lists with ClickHouse repository
sudo apt-get update

# Install ClickHouse server and client version <SERVER_IP>
sudo apt-get install -y clickhouse-server=25.9.3.48 clickhouse-client=25.9.3.48

# Create custom ClickHouse configuration file
sudo tee /etc/clickhouse-server/config.d/benchmark.xml > /dev/null << 'EOF'
<clickhouse>
    <listen_host>::</listen_host>
    <path>/data/clickhouse</path>
    <max_server_memory_usage>106897735680</max_server_memory_usage>
    <max_concurrent_queries>8</max_concurrent_queries>
    <background_pool_size>16</background_pool_size>
    <background_schedule_pool_size>16</background_schedule_pool_size>
    <max_table_size_to_drop>50000000000</max_table_size_to_drop>
</clickhouse>
EOF

# Configure ClickHouse user profile with password and query settings
sudo tee /etc/clickhouse-server/users.d/benchmark.xml > /dev/null << 'EOF'
<clickhouse>
    <users>
        <default replace="true">
            <password_sha256_hex>2cca9d8714615f4132390a3db9296d39ec051b3faff87be7ea5f7fe0e2de14c9</password_sha256_hex>
            <networks>
                <ip>::/0</ip>
            </networks>
        </default>
    </users>
    <profiles>
        <default>
            <max_threads>16</max_threads>
            <max_memory_usage>45000000000</max_memory_usage>
            <max_bytes_before_external_sort>73492193280</max_bytes_before_external_sort>
            <max_bytes_before_external_group_by>73492193280</max_bytes_before_external_group_by>
            <join_use_nulls>1</join_use_nulls>
            <allow_experimental_correlated_subqueries>1</allow_experimental_correlated_subqueries>
            <optimize_read_in_order>1</optimize_read_in_order>
            <max_insert_threads>8</max_insert_threads>
        </default>
    </profiles>
</clickhouse>
EOF

# Start ClickHouse server service
sudo systemctl start clickhouse-server
# Enable ClickHouse server to start on boot
sudo systemctl enable clickhouse-server

Performance Results

Overall Performance Summary

Selected Query Highlights

The following queries demonstrate the performance characteristics observed during testing:

Analysis & Optimization Opportunities

Based on these initial results, several areas for investigation and potential optimization emerge:

Key Questions

• What specific query execution patterns lead to the largest performance differences?
• How do different query types (aggregations, joins, scans) perform comparatively?
• What configuration changes could narrow the performance gap?
• Are there workload-specific tuning opportunities that weren't explored in this initial assessment?

Download Data