Cisco Support Community

Overutilized CPU and Process Decisions


How does a router or switch decide which processes to give precedence to? For example, if a router is facing a DoS attack (or whatever reason) and because of that, its CPU is pegged, how will it decide whether to process a BGP Hello or to policy-switch a data packet with a QoS marking of EF?

In fact, even during normal operatrion, how are the processes prioritized?

I don't think I have ever read a document that addressed this...

Would love some insight.

Thank you



Re: Overutilized CPU and Process Decisions

Many platforms implement a selective packet discard mechanism. Check out the link here:

This helps to protect and preserve the control plane and ensure very high priority packets (RP hellos and such) get priority.

Aside from that, there is also QoS, which ensures, for example, that a parket marked EF would get forwarded before one in the default class.

Super Bronze

Re: Overutilized CPU and Process Decisions

Victor, of course only Cisco can answer your question, but I suspect often the design is there is no inherent priorization for, or between, many processes. The expectation appears to be there will normally be excess CPU capacity so don't bog down the design with unnecessary logic. Confirmation of this might be seen by how badly many Cisco devices perform if you drive the CPU to 100% and keep it there. (BTW: if this is the design, nothing at all wrong with it for a special device such as a router or switch.)


Re: Overutilized CPU and Process Decisions

Thank you, gentlemen.


CreatePlease to create content