Benchmark with pgbench and sysbench

Benchmarking your VPS provides objective data for capacity planning, configuration validation, and provider comparison. pgbench (for PostgreSQL) and sysbench (for MySQL, CPU, memory, and disk I/O) are the standard tools for database and system benchmarking. This guide covers running meaningful benchmarks and interpreting results correctly.

sysbench Installation

# Ubuntu/Debian
sudo apt install sysbench

# RHEL/AlmaLinux
sudo dnf install sysbench

# Verify
sysbench --version

CPU Benchmarking

# Single-threaded CPU benchmark
sysbench cpu --cpu-max-prime=20000 run

# Multi-threaded (use all cores)
sysbench cpu --cpu-max-prime=20000 --threads=$(nproc) run

# Key metric: events per second (higher is better)
# Typical results:
# 1 vCPU VPS: ~800-1200 events/s
# 4 vCPU VPS: ~3200-5000 events/s
# Dedicated: ~5000-8000 events/s per core

Memory Benchmarking

# Sequential memory read/write
sysbench memory --memory-block-size=1M --memory-total-size=10G run

# Random access pattern (more realistic for databases)
sysbench memory --memory-block-size=4K --memory-total-size=10G --memory-access-mode=rnd run

# Key metrics: throughput (MiB/sec) and latency
# Good VPS: >10 GiB/s sequential, >2 GiB/s random

Disk I/O Benchmarking

# Prepare test files (16GB dataset)
sysbench fileio --file-total-size=16G prepare

# Random read/write (database-like workload)
sysbench fileio --file-total-size=16G --file-test-mode=rndrw \
    --time=120 --max-requests=0 --file-extra-flags=direct \
    --threads=4 run

# Sequential read (backup-like workload)
sysbench fileio --file-total-size=16G --file-test-mode=seqrd \
    --time=60 --max-requests=0 --threads=1 run

# Random write (INSERT-heavy workload)
sysbench fileio --file-total-size=16G --file-test-mode=rndwr \
    --time=120 --max-requests=0 --file-extra-flags=direct \
    --threads=4 run

# Cleanup
sysbench fileio --file-total-size=16G cleanup

# Key metrics:
# NVMe VPS: >50,000 IOPS random read, >20,000 IOPS random write
# SSD VPS: >10,000 IOPS random read, >5,000 IOPS random write
# HDD: ~200 IOPS (avoid for databases)

MySQL Benchmarking with sysbench

# Create test database
mysql -u root -p -e "CREATE DATABASE sbtest;"
mysql -u root -p -e "GRANT ALL ON sbtest.* TO 'sbtest'@'localhost' IDENTIFIED BY 'password';"

# Prepare test data (10 tables, 1M rows each)
sysbench /usr/share/sysbench/oltp_read_write.lua \
    --mysql-host=127.0.0.1 --mysql-user=sbtest --mysql-password=password \
    --mysql-db=sbtest --tables=10 --table-size=1000000 \
    prepare

# Read-heavy benchmark (80% reads, 20% writes)
sysbench /usr/share/sysbench/oltp_read_write.lua \
    --mysql-host=127.0.0.1 --mysql-user=sbtest --mysql-password=password \
    --mysql-db=sbtest --tables=10 --table-size=1000000 \
    --threads=16 --time=300 --report-interval=10 \
    run

# Read-only benchmark
sysbench /usr/share/sysbench/oltp_read_only.lua \
    --mysql-host=127.0.0.1 --mysql-user=sbtest --mysql-password=password \
    --mysql-db=sbtest --tables=10 --table-size=1000000 \
    --threads=16 --time=300 \
    run

# Write-only benchmark
sysbench /usr/share/sysbench/oltp_write_only.lua \
    --mysql-host=127.0.0.1 --mysql-user=sbtest --mysql-password=password \
    --mysql-db=sbtest --tables=10 --table-size=1000000 \
    --threads=16 --time=300 \
    run

# Cleanup
sysbench /usr/share/sysbench/oltp_read_write.lua \
    --mysql-host=127.0.0.1 --mysql-user=sbtest --mysql-password=password \
    --mysql-db=sbtest --tables=10 cleanup

PostgreSQL Benchmarking with pgbench

# pgbench comes with PostgreSQL
# Create test database
createdb pgbench_test

# Initialize with scale factor 50 (~750MB dataset)
pgbench -i -s 50 pgbench_test

# Simple read-write benchmark
pgbench -c 20 -j 4 -T 300 pgbench_test
# -c: concurrent clients
# -j: worker threads
# -T: duration in seconds

# Read-only benchmark
pgbench -c 20 -j 4 -T 300 -S pgbench_test

# Custom benchmark with specific query mix
cat > /tmp/custom.sql &1 | grep "events per second"

echo "Memory:"
sysbench memory --memory-block-size=4K --memory-total-size=5G run 2>&1 | grep "transferred"

echo "Disk:"
sysbench fileio --file-total-size=4G prepare > /dev/null 2>&1
sysbench fileio --file-total-size=4G --file-test-mode=rndrw --time=60 \
    --file-extra-flags=direct --threads=4 run 2>&1 | grep -E "reads/s|writes/s|fsyncs/s"
sysbench fileio --file-total-size=4G cleanup > /dev/null 2>&1

echo ""
echo "Apply your tuning changes, then run again to compare."

Automated Benchmark Report

#!/bin/bash
# Generate a complete server benchmark report
REPORT="/tmp/benchmark-report-$(date +%Y%m%d).txt"

{
    echo "=== Server Benchmark Report ==="
    echo "Date: $(date)"
    echo "Hostname: $(hostname)"
    echo "CPU: $(grep 'model name' /proc/cpuinfo | head -1 | cut -d: -f2 | xargs)"
    echo "Cores: $(nproc)"
    echo "RAM: $(free -h | awk '/Mem:/ {print $2}')"
    echo "Disk: $(df -h / | awk 'NR==2 {print $2}')"
    echo ""

    echo "--- CPU Benchmark ---"
    sysbench cpu --cpu-max-prime=20000 --threads=$(nproc) run 2>&1

    echo "--- Memory Benchmark ---"
    sysbench memory --memory-block-size=1M --memory-total-size=10G run 2>&1

    echo "--- Disk I/O Benchmark ---"
    sysbench fileio --file-total-size=4G prepare > /dev/null 2>&1
    sysbench fileio --file-total-size=4G --file-test-mode=rndrw --time=60 --file-extra-flags=direct --threads=4 run 2>&1
    sysbench fileio --file-total-size=4G cleanup > /dev/null 2>&1
} > "$REPORT"

echo "Report saved to $REPORT"

Summary

Benchmarking provides the objective data needed for performance tuning decisions. Use sysbench for system-level metrics (CPU, memory, disk I/O) and MySQL benchmarks, and pgbench for PostgreSQL workloads. Always benchmark before and after configuration changes to measure actual impact. Run benchmarks for at least 5 minutes to get statistically stable results, and pay attention to 95th percentile latency rather than averages — that is what your users experience during peak load.