CPU utilization in a network is an essential parameter that has to be monitored very closely. This is due to the fact that an overloaded CPU will result in switches not responding to OpenFlow requests which in turn leads to LLDP timeout. These events make the controller believe there are link-flaps present in the network, leading to distorted networks.

There are four events that utilize the maximum resources of a switch CPU, and each requires unique isolation mechanisms:

• New flow messages
• Controller requests
• Slow-path packet forwarding
• Internal state keeping