As of the Linux 2.6 kernel, the default I/O Scheduler is Completely Fair Queuing (CFQ). The scheduler is an effective solution for nearly all workloads.
With the release of Linux 2.6, the kernel has these schedulers as configurable options:
Completely Fair Queuing (cfq): CFQ is the default under many Linux distributions. CFQ places synchronous requests submitted by processes into a number of per-process queues and then allocates time slices for each of the queues to access the disk. The length of the time slice and the number of requests a queue is allowed to submit depends on the I/O priority of the given process. Asynchronous requests for all processes are batched together in fewer queues, one per priority.
NOOP (noop): NOOP is the simplest I/O scheduler for the Linux kernel based upon the FIFO queue concept. The NOOP scheduler inserts all incoming I/O requests into a simple FIFO queue and implements request merging. The scheduler assumes I/O performance optimization will be handled at some other layer of the I/O hierarchy.
Anticipatory (anticipatory): Anticipatory is an algorithm for scheduling hard disk I/O. It seeks to increase the efficiency of disk utilization by "anticipating" synchronous read operations.
Deadline (deadline): The goal of the Deadline scheduler is to guarantee a start service time for a request. It does this by imposing a deadline on all I/O operations to prevent starvation of requests.
The default scheduler will affect all disk I/O for VMDK and RDM-based virtual storage solutions. In virtualized environments, it is often not beneficial to schedule I/O at both the host and guest layers. If multiple guests use storage on a filesystem or block device managed by the host operating system, the host may be able to schedule I/O more efficiently because it is aware of requests from all guests and knows the physical layout of storage, which may not map linearly to the guests' virtual storage.
Testing has shown that NOOP or Deadline perform better for virtualized Linux guests. ESX uses an asynchronous intelligent I/O scheduler, and for this reason virtual guests should see improved performance by allowing ESX to handle I/O scheduling.
To implement this change, please refer to the documentation for your Linux distribution.
Note: All scheduler tuning should be tested under normal operating conditions as synthetic benchmarks typically do not accurately compare performance of systems using shared resources in virtual environments.
For example, this change can be implemented by:
The scheduler can be set for each hard disk unit. To check which scheduler is being used for particular drive, run this command:
For example, to check the current I/O scheduler for sda:
# cat /sys/block/sda/queue/scheduler
[noop] anticipatory deadline cfq
In this example, the sda drive scheduler is set to NOOP.
To change the scheduler on a running system, run this command:
# echo scheduler > /sys/block/disk/queue/scheduler
For example, to set the sda I/O scheduler to NOOP:
# echo noop > /sys/block/sda/queue/scheduler
Note: This command will not change the scheduler permanently. The scheduler will be reset to the default on reboot. To make the system use a specific scheduler by default, add an elevator parameter to the default kernel entry in the GRUB boot loader menu.lst file.
For example, to make NOOP the default scheduler for the system, the /boot/grub/menu.lst kernel entry would look like this:
title CentOS (2.6.18-128.4.1.el5)
kernel /vmlinuz-2.6.18-128.4.1.el5 ro root=/dev/VolGroup00/LogVol00 elevator=noop
initrd /initrd-2.6.18-128.4.1.el5.imgWith the elevator parameter in place, the system will set the I/O scheduler to the one specified on every boot.