Disk performance monitoring

The following performance counters can be used to monitor disk I/O activities:

Counter Name	Description
Avg. Disk sec/Read	Shows the average time, in seconds, of a read of data from the disk.
Avg. Disk sec/Write	Shows the average time, in seconds, of a write of data to the disk.
Disk Transfers/sec	Shows the rate of read and write operations on the disk.
Disk Bytes/sec	Shows the rate at which bytes are transferred to or from the disk during write or read operations.
Avg. Disk Queue Length	Shows the average number of both read and write requests that were queued for the selected disk during the sample interval.

• PhysicalDiskThe related performance counters are mainlyPhysical diskIf multiple partitions exist on the same physical disk, the counter calculates the sum of all partitions.
• LogicalDiskThe related performance counters are mainlyLogical PartitionInstead of monitoring the entire disk.

Where,Avg. Disk sec/ReadAndAvg. Disk sec/WriteIt is two counters that mainly consider disk I/O performance. They do not need additional information for reference and can directly determine the average latency of I/O requests.

Avg. Disk sec/Read

Describes the average time (in seconds) required to read data from a disk ).

Avg. Disk Sec/Read is the average time in seconds of a read of data from the disk.

The meanings of the parameter value range are listed below:

Range	Represent
Less than 10 MS	Good
Between 10-20 MS	Yes
Between 20-50 MS	A little slow, please note
More than 50 milliseconds	Severe I/O bottlenecks

The numbers listed here are for general reference. For applications with special requirements, there are higher requirements for disk parameters.

Avg. Disk sec/Write

Describes the average time (in seconds) required to write data from a disk ).

Avg. Disk Sec/Write is the average time in seconds of a write of data to the disk.

The meanings of the parameter value range are listed below:

Range	Represent
Less than 10 MS	Good
Between 10-20 MS	Yes
Between 20-50 MS	A little slow, please note
More than 50 milliseconds	Severe I/O bottlenecks

In fact, the parameter scope andAvg. Disk sec/ReadIs similar.

If latency is high for most disk drives, bottlenecks may occur in Communication media (Communication Medium), such as san hba, switch, optical fiber, front-end adapter CPU and cache. If there is only one or a few latencies on the drive, the bottleneck is usually the disk group (JBOD: Just a Bunch of Disks ), for further investigation, you need to view the values of other counters on the specified disk drive.

JBOD (for "just a bunch of disks," or sometimes "just a bunch of drives ") is an array of hard disks that haven't been configured according to the RAID (redundant array of independent disks) system.

Disk Transfers/sec

Disk Bytes/secAndDisk Transfers/secThe two performance counters are often used to observe the scale and quantity of I/O requests, to determine whether it is a JBOD problem or a communication medium problem, and then passAvg. Disk Queue LengthCounter to determine whether it is a communication medium problem.

Disk Transfers/secIt is actually causedDisk Reads/secAndDisk Writes/secComposed of two counters.

Disk Transfers/sec = Disk Reads/sec + Disk Writes/sec

The value of the counter above can be used to determine whether the drive does not have sufficient disk support to determine whether the RAID on a given number of disks is reasonable. You can calculate the RAID type based on the formula listed in the following table.

RAID Type	I/OS per Disk
RAID 0	(Reads + writes)/number of disks
RAID 1	[Reads + (2 * writes)]/2
RAID 5	[Reads + (4 * writes)]/number of disks
RAID 10	[Reads + (2 * writes)]/number of disks

For example, a 15000 RPM (Per Minute, Revolutions per Minute) disk can provide about 180 IOPS (I/O requests Per second. IfDisk Transfers/secIf the maximum value is 1800, 1800/180 = 10, then the drive requires at least 10 15000 RPM disks to form a RAID group.

If the latency is caused by the number of disks, consider the following solution:

• Use a faster disk.
• Move frequently accessed files to a separate disk.
• Add more disks to the RAID array.
• Use a faster RAID type.
• Stop sharing disks with other volumes.

If the latency is caused by transmission media problems, consider the following solution:

• Move frequently accessed files to a separate disk.
• Verify SAN's high-speed cache.
• Multiple channels are used.

References

• Monitoring Disk Usage
• Windows Performance Monitor Disk Counters Explained
• Top Six FAQs on Windows 2000 Disk Performance
• Troubleshooting Slow Disk I/O in SQL Server
• PhysicalDisk Object
• Creating a custom performance monitor counter for SQL Server
• How to use the SQLIOSim utility to simulate SQL Server activity on a disk subsystem

In this article, "disk performance monitoring" is published by Dennis Gao from the blog Park. Without the author's consent, reprinting in any form is prohibited. Any automatic or manual crawler reprinting behavior is a rogue.

Disk performance monitoring