Configuring and Optimizing Your I/O Scheduler

Selecting and Configuring Your I/O Scheduler

The default I/O scheduler is selectable at boot time via the iosched kernel command-line parameter. Valid options are as, cfq, deadline, and noop. The I/O scheduler is also runtime-selectable on a per-device basis via /sys/block/device/queue/scheduler, where device  is the block device in question. Reading this file returns the current I/O scheduler; writing one of the valid options to this file sets the I/O scheduler. For example, to set the device hda to the CFQ I/O Scheduler, one would do the following:

  # echo cfq > /sys/block/hda/queue/scheduler

The directory /sys/block/device/queue/iosched contains files that allow the administrator to retrieve and set tunable values related to the I/O scheduler. The exact options depend on the current I/O scheduler. Changing any of these settings requires root privileges.

A good programmer writes programs that are agnostic to the underlying I/O subsystem. Nonetheless, knowledge of this subsystem can surely help one write optimal code.

Optimizing I/O Performance

Because disk I/O is so slow relative to the performance of other components in the system, yet I/O is such an important aspect of modern computing, maximizing I/O performance is crucial.

Minimizing I/O operations (by coalescing many smaller operations into fewer larger operations), performing block-size-aligned I/O, or using user buffering (see Chapter 3), and taking advantage of advanced I/O techniques, such as vectored I/O, positional I/O (see Chapter 2), and asynchronous I/O, are important steps to always consider when system programming.

The most demanding mission-critical and I/O-intense applications, however, can employ additional tricks to maximize performance. Although the Linux kernel, as discussed previously, utilizes advanced I/O schedulers to minimize dreaded disk seeks, user-space applications can work toward the same end, in a similar fashion, to further improve performance.