Disk Request Service Table

Function:

• To display disk request time statistics.

Description:

• This table presents the disk request times.

Name - The name of the raw disk drive device file.

I/O - The number of requests to this disk per second.

Request - The average time to service disk requests.

Service - The average time taken to service requests including wait time in the queue.

Active - The amount of time the disk was busy as a percentage.

The Disk Request View profiles the disk queues and provides an insight into the efficiency of the disk subsystem. Ideally, request and service times should be as low as possible. If the service time is significantly longer than the request time, it indicates that the disk requests are waiting in the disk queue too long before being serviced by the disk driver. This may be due to excessive I/O load in which case a faster disk controller, disk drives or both may be needed. Long queue wait times can also be caused by too small a buffer cache or a fragmented disk. If the buffer cache is too small, normal application load will translate into high disk I/O request rates. See the System Buffering View to investigate this further.

A disk is said to be fragmented when logically sequential blocks are physically stored on the disk in non-contiguous regions. Blocks of large files are then scattered all over the disk, causing the disk arm to move randomly when reading or writing the required blocks.

Filesystems are badly ordered when their positioning causes disk accesses to be spaced widely over the disk. If the two most commonly used filesystems are at opposite ends of a disk, performance will be sub-optimal.

If filesystem performance is slow, you may have a fragmented disk or poorly ordered filesystems. To see if this is the case, compare the request time displayed in this view with the average access time displayed in the Disk Configuration View. If the request time is considerably longer than the average access time, you have a disk access problem. The request time increases because the disk driver is servicing I/O requests for blocks which are widely scattered over the disk causing the driver to spend extra time locating blocks and moving the disk heads for each request.

Use the File System Balance View to determine which are the most heavily used filesystems. You will increase I/O performance by moving these filesystems together and to the front of the disk.

If you suspect a disk fragmentation bottleneck, you can save and restore the filesystem to tape using a file by file backup. This will ensure logically contiguous data is also physically contiguous.