Home > Workload Solutions > Oracle > Best Practices > AMD-Based Oracle Best Practices on Dell PowerEdge R740 and PowerMax 2000 > Linux Best Practices > Day Three Best Practices > Red Hat Enterprise Linux: Disk Schedulers
In this best practice, we changed the default Linux disk scheduler per Red Hat guidelines.
Category | Operating System |
Product | Red Hat Enterprise Linux 8.3 |
Type of best practice | Performance Optimization |
Day and value | Day 3, Fine Tuning |
Overview
There are several choices with Red Hat Enterprise Linux 8 for the type of disk scheduler used by the operating system. The function of the disk schedule is for ordering I/O requests to storage. Each disk scheduler provides a different method for managing storage requests:
In our best practice, we followed Red Hat’s recommendation of using the none disk scheduler as the PowerMax used NVMe drives.
Recommendation
The none disk scheduler provided a slight performance increase for the following metrics:
The performance metrics below showed no significant changes:
Overall, the None disk scheduler increased storage related performance as reflected in the increase in PowerMax IOPS, NOPM, and TPM metrics. There was a very slight drop in write latency on the PowerMax storage array and Log File Parallel Writes which should have improved our IOPS, NOPM and TPM metrics. Although the very slight decreases in write times were not significant the impact over time can increase performance. We believe this is what was observed in our OLTP load test.
Implementation Steps
To set disk scheduler using TUND:
$ udevadm info --query=property --name=/dev/device | grep -E '(WWN|SERIAL)'
[main] Include= existing-profile [disk] devices_udev_regex=ID_WWN=device system unique id elevator=selected-scheduler
devices_udev_regex=(ID_WWN=0x5002538d00000000)|(ID_WWN=0x1234567800000000) elevator=none
# tuned-adm profile my-profile
$ tuned-adm active $ tuned-adm verify
Additional Resources
Red Hat Enterprise Linux 8: Setting the disk scheduler
Red Hat Enterprise Linux 8: Monitoring and managing system status and performance