The following figure shows an example of a network, called Network A, in which each adapter is connected to only one port on the storage system. The host bus adapters are connected directly to the storage system ports and are not connected via hubs or switches.

In Network A, ports 1A and 1C are connected to high-priority production servers. Port 2A is connected to a low-priority development server. Therefore, high priority (Prio.) is set to ports 1A and 1C, and low priority (Non-Prio.) is set to port 2A. The following figure shows the priority settings for these ports on the Port tab of the Server Priority Manager main window.

The I/O traffic at the ports is now monitored. When the desired amount of data has been collected, the traffic statistics are reviewed on graphs using the Performance Monitor window. The following figure shows the changes in the I/O rate for the three ports (1A, 1C, and 2A). According to the data, the I/O rate for ports 1A and 1C was approximately 400 IO/s at first, and the I/O rate for port 2A was approximately 100 IO/s at first. However, as the I/O rate for port 2A gradually increased from 100 IO/s to 200 IO/s, the I/O rate for ports 1A and 1C decreased from 400 IO/s to 200 IO/s, which indicates that the performance of the high-priority production servers decreased. If you want to maintain the I/O rate for prioritized ports 1A and 1C at 400 IO/s, you can set an upper limit to the I/O rate for the non-prioritized port 2A.

When an upper limit is set for the first time, it is recommended that the upper limit be approximately 90 percent of the peak traffic. In Network A, the peak I/O rate for the non-prioritized port (2A) is 200 IO/s, so the recommended upper limit for port 2A is 180 IO/s. You can then adjust this value later as needed according to the results of your data analysis.

After upper limit values have been applied, the traffic at the ports is measured to check the result of the upper limits and see whether host performance has improved to a desirable level, in this case 400 IO/s for ports 1A and 1C. If production server performance is not improved to a desirable level, you can adjust the upper limit to a smaller value until the I/O rate for the high priority servers reaches the desirable level.

If you want to allow improved performance for non-prioritized ports when system activity is low, you can apply a threshold. For one-to-one connections threshold values can be set in either of the following two ways:

• Set a threshold to each prioritized port.

  In Network A, if you set a threshold of 200 IO/s to port 1A and a threshold of 100 IO/s to port 1C, the upper limit on the non-prioritized port (2A) is disabled when both of the following conditions occur:

  The I/O rate for port 1A is 200 IO/s or lower.

  The I/O rate for port 1C is 100 IO/s or lower.

• Set one threshold to the entire storage system.

  In Network A, if you set a threshold of 500 IO/s to the storage system, the upper limit on the non-prioritized port (2A) is disabled when the sum of the I/O rates for all prioritized ports (1A and 1C) falls below 500 IO/s.

The following table shows the relationship between the thresholds of prioritized ports and the upper limits of non-prioritized ports.

Threshold settings	Upper limit is set (non-zero) on the non-prioritized port	Upper limit of the non-prioritized port = 0
Threshold is set to the prioritized port	When thresholds are set on multiple prioritized ports, depending on the transfer rate, the following controls are executed: When the I/O rate or transfer rate goes above the threshold in any prioritized port, the upper limits of all non-prioritized ports are in effect. When the I/O rate or transfer rate goes below the threshold in all prioritized ports, the upper limits of all non-prioritized ports are not in effect.	The threshold control of the prioritized port is not executed.
Threshold is not set to the prioritized port	The specified upper limit is always in effect.