> On Fri, 11 Jul 1997, Robert Hiltibidal wrote:
>
>
> > There are other far more simpler ways to check a network component that a
> > cpu monitor.
>
> Really? If you have an alternate method of determining CPU load given X
> number of BGP views, full or partial, Y numbers of filters, and Z, where Z
> is variable traffic load, I am all ears.
Agreed. But lets expand the list of things that eat CPU:
- snmp: How much load does a snmpwalk over ethernet generate?
- admin telnet & pmconsole: I know that "erase all-flash" causes all PM
fuctions to freeze for a few seconds
- ICMP: echo-request, destination-unreachable (for traceroute), etc
- routing updates: Both BGP and OSPF use CPU based on how much the
table changes. If it changes a lot, it will eat a lot of CPU.
- size of routing table: the bigger the table, the more CPU it takes to
find the proper destination
Tom