This article does not apply to unmanaged switches.
The management interface of your switch relies on the CPU of the switch to respond to ICMP ping requests. Such requests are given a low priority by the switch CPU relative to other tasks. This means that you may see a delayed or even no response when pinging the IP address of the switch if the switch is busy with other tasks.
It is important to understand that the switch CPU is not involved in the forwarding of traffic through the switch. As such, the ICMP response time from the switch management interface and, therefore the switch CPU, should not be used as a measurement of the overall performance of the switch.
The following considerations should be taken into account when implementing monitoring of a switch:
- Where possible, it is recommended to use SNMP to monitor the switch instead of ICMP. You can use the NMS300 software to monitor all NETGEAR switches that support SNMP.
- If ICMP is required in your environment, you should ping devices through the switch as opposed to pinging the switch directly. This means pinging a device connected to the switch from a station connected to the same switch.
- If you still rely on pinging the management interface of the switch, consider adding a delay in order to prevent false down alerts. This will allow a moderate load on the switch CPU to not result in false alerts. Consider providing a minimum of a 2 second interval between each ICMP message and implement a minimum of 10 retries before triggering a down alert.
- By design, it is not possible to alter the priority of ICMP messages sent to the CPU. However, if your model of switch supports DiffServ, then you can create a DiffServ class and policy to filter ICMP traffic from a specific source address (e.g. a monitoring server) and assign it to a higher queue in order to give it more priority over other traffic. This will only help in a situation where the switch is running at high but not peak CPU usage. This involves the following steps:
- Create a DiffServ class and apply a source IP or subnet of a system that needs to send high priority traffic as rule to this class. Limit to a specific protocol in order to prevent all traffic from getting high queue.
- Create a policy for this class and assign the traffic to highest queue.
- Apply the policy to an inbound port from which the source IP or subnet will reach the switch management IP address.
- If the switch is running at peak or high CPU usage, there is a possibility that some functions including ICMP responses from the management interface will be degraded to some extent. In such a situation, it is important to first determine if there are abnormal factors in the network which may be impacting on the switch and leading to a high CPU load.