Re: [Pce] Whither Stateless PCE?

[Robert Varga <[e-mail]>]

On 05/09/2016 03:06 PM, [e-mail] wrote:
> Hi Xian,
> 
>  
> 
> Regarding the METRIC object, the issue is not having the Object in the
> PCRpt (which already works). The issue is that the metric object will
> reflect the operational state of the LSP rather than it’s configuration.
> 
> The best example may be :
> 
> PCC is configured to use IGP metric with a cost bound to 20.
> 
> In the PCRpt, it will send in the METRIC object the operational values
> of the LSP, so the metric may be anything below 20 (e.g. 14) and B=0 and
> we will not have any information about the cost boundary constraint, so
> the PCE will not be able to fullfill this constraint when computing the
> path for the PCC.

Hello,

This looks like the 'ERO/RRO in PCRpt' discussion we have had with
Julien. While the ERO/RRO split makes this a non-issue in that case, we
seem to lack a general mechanism to discern 'intended' and 'effective'
state, notably in the case when the PCE needs to recover the intended
state from the PCC.

Does that sum up the problem?

Thanks,
Robert