question Re: some clarification about the explanation regarding 'tuned', such as its potential impact on Hadoop. in Support Questions

some clarification about the explanation regarding 'tuned', such as its potential impact on Hadoop.

steinsgate — Tue, 21 Apr 2026 06:32:35 GMT

Hi everyone, I would like some clarification. Thank you.

CDP recommends disabling the 'tuned' service for performance optimization.

The suggestion to disable 'tuned' stems from its capability to transition the CPU into c-state idle mode, potentially leading to decreased performance.

Disable the tuned Service on RHEL 8

We would like some clarification about the explanation regarding 'tuned', such as its potential impact on Hadoop. Thank you.

For example, THP:

Disabling Transparent Hugepages (THP)
Most Linux platforms supported by Cloudera Runtime include a feature called transparent hugepages, which interacts poorly with Hadoop workloads and can seriously degrade performance.

For example, swappiness:

Setting the vm.swappiness Linux Kernel Parameter

On most systems, vm.swappiness is set to 60 by default. This is not suitable for Hadoop clusters because processes are sometimes swapped even when enough memory is available.

Regards

Re: some clarification about the explanation regarding 'tuned', such as its potential impact on Hadoop.

upadhyayk04 — Mon, 22 Apr 2024 15:50:58 GMT

Hello @steinsgate

Thank you for reaching out on this

You can refer to the below RedHat link to know more about how tuned works on OS. You can check from the below document how different classes work

https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/monitoring_and_managing_system_status_and_performance/getting-started-with-tuned_monitoring-and-managing-system-status-and-performance

You can set the particular workloads from the available ones using tuned.

Also tuned is providing a set of guidelines. You still need to do testing for your particular workloads to validate that the performance curve is acceptable.

Re: some clarification about the explanation regarding 'tuned', such as its potential impact on Hadoop.

steinsgate — Tue, 23 Apr 2024 08:02:06 GMT

Hello @upadhyayk04

I see, thank you.