Hi Mach,

>>> So, I think that PCEP needs to be able to indicate
>>> the sequence of domains that should be traversed.
>>
>> The easy way to handle this is by using the IRO
>> object specified in PCEP. I can add:
>>
>> "The sequence of domains to be traversed MAY be
>>  encoded in the IRO so that it is available to all
>> PCEs."
>
> As Adrian pointed out:"A PCE may serve more than one
> domain." , and I am not sure whether there are multiple
> PCEes may serve one domain (for HA or load balance reason).

Absolutely. A domain may have multiple PCEs (for example, one at each domain 
border router).

> If my assumption is reasonable, I am afriad just using the
> IRO is not the best idea. Becasue, just based on the IRO,
> for a specific PCE, how does it accurately know which
> domain(s) it should serve?
> For example:there are three domains(d1,d2,d3) and three PCEs
> (PCE1,PCE2,PCE3), PCE1 serves d1, PCE2 could serve d2 and d3,
> and PCE3 also could serve d2 and d3, for such condition, how
> does BRPC work?

There are two questions to be answered by a PCE:
- which is the previous/next domain in the path?
- which is the next PCE to talk to?

We now have mechanisms to convey both pieces of information as follows:
- Use the IRO to include a list of domain identifiers
  to indicate the sequence of domains to be traversed.
  In your example, the PCReq from the original PCC (or
  maybe from the first PCE) would have an IRO showing
  the domain sequence d1, d2, d3
- Use the list of PCE-ID objects as described in
  draft-ietf-pce-monitoring-01.txt. In your example the
  sequence might be PCE1, PCE3.

The inclusion of these pieces of information would be optional. Failure to 
include sufficient information to enable path computation will (obviously) 
result in a failure.

In the BRPC case you mention, the IRO would need to be propagated from the 
source PCE to the final PCE as the PCReq is forwarded. The sequence of PCEs 
through which the PCReq is forwarded, can be determined either from the IRO 
or from the PCE-ID objects.

Hope this helps,
Adrian 


_______________________________________________
Pce mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/pce

Reply via email to