Especially with the addition of the routing protocols OSPF and BGP it is
*very* important to be able to watch CPU usage.
Here is what a Cisco shows:
CPU utilization for five seconds: 8%/4%; one minute: 8%; five minutes: 7%
PID QTy PC Runtime (ms) Invoked uSecs Stacks TTY Process
1 ME 322DBEE 55348 8259 6701 1544/3000 0 OSPF Router
2 Lst 30B4362 758456 9165 82755 1800/2000 0 Check heaps
3 Mst 30DACDE 76400 3370852 22 844/2000 0 Timers
4 Lwe 31318A2 20012 25833 774 1432/2000 0 ARP Input
5 Lwe 314AA9E 0 1 0 1768/2000 0 Probe Input
6 Mwe 314A41A 0 1 0 1784/2000 0 RARP Input
7 Hrd 313C738 40922416 22950204 1783 2728/4000 0 IP Input
8 Mwe 3164404 6000 107009 56 1392/2000 0 TCP Timer
..etc..
On Fri, 11 Jul 1997, Jon Lewis wrote:
> Sure...everything that happens on such a system affects the CPU. I just
> don't buy the excuse "it would take a major rewrite to monitor CPU use
> in ComOS". If it's a time sharing system, just write a CPU montoring
> process, and see how much time it gets to share :) It can't be as bad
> as they make it sound. And like others have said, would you trust a BGP
> router if you could have no idea how loaded the CPU is? When the router
> starts acting up, is it bugs? is it the CPU falling over? Who knows?