Martin,
As long as CPU utilization is not affecting forwarding performance,
I can live with it. But what I don't understand (don't like) is that
IOS scheduler is preempting (or not giving CPU fast enough) to
whatever process is responsible for forwarding.
In my case, it looks like MPLS frame removal is done in user-land
(is there anything like this in IOS?), so it depends on CPU being given
to some process. I would expect, that such operation should be done
during interrupt processing, not by regular process...
Przemek
On Fri, 2002-07-12 at 16:06, martin.laub@attla.net.ar wrote:
> Hi,
>
> I suffer the same problem with 12.0(21)ST with tag switching disabled
>
> The delay in RTT increases every 1 minute, at the same time BGP process scans the routing table. There´s an explanation at http://www.cisco.com/warp/public/459/highcpu-bgp.html
>
> Rgds.
>
> ML
>
>
> >All,
> >
> >I have recently encountered interesting phenomenon:
> >
> >On cisco 7206 working in a typical configuration in ISP >environment (BGP, ISIS, MPLS), I have noticed that traffic forwarded from ATM interface (PA-A3-OC3) to T1 interface (PA-8CT1) is prone to increased forwarding latency (up to 700ms!) while CPU utilization
> is high (during typical spikes while BGP scenner or ISIS SPF runs).
>
> This is happening only when forwarded packets are arriving as MPLS,
> and egressing as normal IP HDLC frames. In other words, router is
> removing MPLS label (penultimate hop).
>
> As soon as I disabled tag switching on ATM, problem is gone.
>
> IOS version is 12.2(8)T4.
>
> Is it a feature or bug?
>
> I can replicate it easily.
>
> Thanks,
>
> Przemek
>
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:04 EDT