[Pce] RE: Objective functions in PCECP (draft-ietf-pce-comm-protocol-gen-reqs-04.txt)

Fri, 17 Feb 2006 07:42:17 -0800

Hi Filippo,
 
Please see inline,

De : Filippo Cugini [mailto:[EMAIL PROTECTED]
Envoyé : jeudi 16 février 2006 19:55
À : LE ROUX Jean-Louis RD-CORE-LAN
Cc : [EMAIL PROTECTED]; [EMAIL PROTECTED]
Objet : Re: Objective functions in PCECP (draft-ietf-pce-comm-protocol-gen-reqs-04.txt)

Hi,
 

JLLR> This depends on the application. For some applications you may want a path with a minimum delay..

I absolutely agree. Indeed delay (like bandwidth..) is a QoS requirment and could be different for every request.

However, in section 6.1.17 it is mentioned for example:

"Maximize the residual bandwidth on the most loaded link"
that deals with Traffic Engineering optimization, and it not related to QoS application requirements 
 
One may want to sent a request message including a set of path request, to be computed in a synchronized manner,  with as objective to maximize the residual bandwidth...
 
JLLR> The PCE can reject a request.. 
Agree, however, do you think that some objective functions indicated in section 6.1.17 will be ever accepted by a PCE belonging to a different domain?  
 
When domains are areas, sub-ASs or ASs of the same provider, then the answser is YES...
When domains are ASs of distinct network providers, then it will depend on metric normalization aspects and on the level of trust between network providers, and I would say yes, at least for the shortest path and max bandwidth path. 
Anyway, note that this is a generic requirement draft, not related to a specific PCE application.
 
Best Regards,
 
JL
 
 
Best Regards
 
JL 
 
Best regards
 Filippo  
 
 
----- Original Message -----
Sent: Thursday, February 16, 2006 5:20 PM
Subject: RE: Objective functions in PCECP (draft-ietf-pce-comm-protocol-gen-reqs-04.txt)

Hi Filippo
 
Please see inline

De : Filippo Cugini [mailto:[EMAIL PROTECTED]
Envoyé : jeudi 16 février 2006 16:17
À : LE ROUX Jean-Louis RD-CORE-LAN
Cc : [EMAIL PROTECTED]; [EMAIL PROTECTED]
Objet : Re: Objective functions in PCECP (draft-ietf-pce-comm-protocol-gen-reqs-04.txt)

Hi JL,

 

Ok for the distinction, however, I still think objective functions should not be included in PCECP.

It is not clear to me why PCC should require different objective functions for different requests.  

 

JLLR> This depends on the application. For some applications you may want a path with a minimum delay, for other applications you may want a shortest cost delay bounded path, for other applications you may want a delay bounded widest path (e.g. upon blocking case, if the requested bandwidth cannot be fit one may want the widest path), etc....

 

Moreover, in case of PCE-PCE request, a PCE could affect TE optimization also within other domains. 

 

JLLR> This is up to the PCE to apply local policies and to filter request parameters. The PCE can reject a request if the requested objective function is not allowed.

Also note that inter-AS specific aspects are beyond the scope of this draft. This should be covered in the PCECP inter-AS requirement draft.

 

Best Regards

 

JL

 

Thank you

 

  Filippo

 
 
----- Original Message -----
Sent: Thursday, February 16, 2006 12:13 PM
Subject: RE: Objective functions in PCECP (draft-ietf-pce-comm-protocol-gen-reqs-04.txt)

Hi Filippo
 
Please note that "MAY include" =/= "MUST support the inclusion"...
 
>As a general comment: is it really a must to include in a communication protocol information that may be common for all requests?
>maybe something could be better done through configuration
 
No the inclusion is a MAY, but the protocol MUST allow you to do it if you want...
 
 
Regards,
 
JL

De : Filippo Cugini [mailto:[EMAIL PROTECTED]
Envoyé : mercredi 15 février 2006 15:59
À : [EMAIL PROTECTED]; LE ROUX Jean-Louis RD-CORE-LAN
Cc : [EMAIL PROTECTED]
Objet : Objective functions in PCECP (draft-ietf-pce-comm-protocol-gen-reqs-04.txt)

Hi,
 
In section 3 (Introduction):
[..] The path computation request [..] MAY also include an objective function.
In the rest of the document, MUST is always used instead of MAY
 
One comment about the use of "MUST" related to objective functions.
In my view some information are request-specific (source, dest, bandwidth..),others are general (e.g., objective functions), i.e. they should be the same for every request. For example, what about the overall network resource utilization (i.e. optimization) if PCC requires different objective functions for different requests?
 
As a general comment: is it really a must to include in a communication protocol information that may be common for all requests? maybe something could be better done through configuration
 
Thank you
  Filippo
_______________________________________________
Pce mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pce

Reply via email to