Linux Swaps in ClickHouse: Peak Performance in Memory-Intensive Operations


“Swap wisely, perform optimally: Fine-tuning Linux swap is the silent key to unleashing ClickHouse’s full potential.”

Excessive Linux swaps can significantly degrade the performance of memory-intensive applications, including databases like ClickHouse. When the system starts swapping out memory pages to disk, it effectively substitutes fast RAM operations with much slower disk I/O operations. Given that ClickHouse is designed to perform many operations in memory for speed, relying on swap can seriously undermine its performance.

Impact of excessive Linux Swaps on ClickHouse

  1. Latency: Disk operations are orders of magnitude slower than RAM operations. When ClickHouse’s working set is moved to swap, query response times increase.
  2. Throughput: The overall number of queries ClickHouse can handle in a given time frame decreases.
  3. CPU Overhead: Swapping introduces additional CPU overhead due to the management of page tables and the actual I/O operations.
  4. Disk I/O Saturation: Excessive swapping can saturate disk I/O, affecting other I/O-bound operations like data writes or merges in ClickHouse.

Tips and Tricks to Optimize Linux Swapping for ClickHouse

  1. Limit Swappiness
    • Tip: The vm.swappiness parameter controls how aggressively Linux swaps out memory. A higher value makes the system more prone to swap.
    • Trick: Set vm.swappiness to a low value (e.g., 10 or even 1) to reduce swapping:
sysctl -w vm.swappiness=10
  1. Adequate RAM
    • Tip: Ensure the server running ClickHouse has enough RAM to accommodate its workload.
    • Trick: Monitor RAM usage regularly. If consistently high, consider adding more RAM to the server.
  2. Monitor Swap Usage
    • Tip: Regularly check the amount of swap space used.
    • Trick: Use commands like free -h or vmstat to keep an eye on swap usage. If swap usage is consistently high, it’s a sign that you need to adjust your setup.
  3. Use Faster Swap Devices
    • Tip: If some swapping is inevitable, consider using faster devices for swap.
    • Trick: Use SSDs for swap space instead of HDDs. Even better, consider NVMe SSDs for their lower latency.
  4. Limit ClickHouse Memory Usage
    • Tip: ClickHouse settings can be tweaked to limit its memory footprint.
    • Trick: Adjust ClickHouse settings like max_memory_usage or max_bytes_before_external_group_by to ensure it doesn’t use more memory than what’s available.
  5. Separate Swap and Data Disks
    • Tip: If ClickHouse data and swap are on the same disk, they can compete for I/O, further degrading performance.
    • Trick: Ensure that the swap partition or file is on a different physical disk than ClickHouse data directories.
  6. Consider Zswap or Zram
    • Tip: These are compressed cache solutions that can help reduce the I/O impact of swapping.
    • Trick: Enable Zswap or Zram if your Linux distribution supports it. They can compress swap data, reducing the amount of actual disk I/O.


While ClickHouse is designed for in-memory operations, ensuring optimal swap configuration on Linux is crucial for those edge cases where it’s needed. By minimizing the reliance on swap and ensuring that it’s as efficient as possible when swapping occurs, you can maintain the high performance that ClickHouse is known for.

To know more about Linux in the context of ClickHouse, do consider reading the following articles:

  1. Python script to monitor Linux Process Memory matrix
  2. Python script to monitor Linux Disk IO matrix
  3. Tuning Linux Transparent Huge Pages (THP) for ClickHouse Performance
About Shiv Iyer 222 Articles
Open Source Database Systems Engineer with a deep understanding of Optimizer Internals, Performance Engineering, Scalability and Data SRE. Shiv currently is the Founder, Investor, Board Member and CEO of multiple Database Systems Infrastructure Operations companies in the Transaction Processing Computing and ColumnStores ecosystem. He is also a frequent speaker in open source software conferences globally.