Re: [Idr] I-D Action: draft-ietf-idr-rpd-05.txt

[Gyan Mishra <[e-mail]>]

Hi Huaimo

Few more questions in-line

Thank you

Gyan

On Tue, Jun 30, 2020 at 8:53 PM Huaimo Chen <[e-mail]>
wrote:

> Hi Gyan,
>
>     Thank you very much for your comments.
>
>     Our answers/explanations are inline below with prefix [HC].
>
> Best Regards,
> Huaimo on behalf of authors
> ------------------------------
>
> *From:* Idr <[e-mail]> on behalf of Gyan Mishra <
> [e-mail]>
> *Sent:* Monday, June 22, 2020 7:15 PM
> *To:* [e-mail] <[e-mail]>
> *Cc:* [e-mail] <[e-mail]>
> *Subject:* Re: [Idr] I-D Action: draft-ietf-idr-rpd-05.txt
>
>
>
>
>
> Authors
>
>
>
> Few comments
>
>
>
> In the abstract and introduction it mentions adjusting traffic.  When
> reading the abstract and introduction adjusting traffic the first thing I
> think of when saying adjusting is BGP QOS QPPB policy to rate limit BGP
> traffic.
>
>
>
> [HC]: It seems that the way (or say RPD) in which the traffic is adjusted
> in the draft and BGP QOS QPPB are different. The BGP QOS QPPB is based on
> configurations or commands. Different configurations for different parts of
> the network need to be done. These configurations combined with the BGP
> propagation provides some QoS policy on some traffic. The RPD provides a
> way to adjust the traffic (i.e., adjust the route attributes to redirect
> the traffic to the less used links) automatically without those
> configurations on different parts of the network.
>

    Gyan> The point here I was trying to make is that in the intro and
draft you mention adjusting traffic to me that means BGP QOS rate limiting
which may also be accomplished by RPD, but I think what may better
represent RPD is stating RPD accomplishes the goal of automated update of
 complex multi active routing policy using a centralized PCE controller
using large communities atoms containers for policy granularity and
flexibility.

https://tools.ietf.org/html/draft-ietf-pce-pcep-extension-native-ip-05


>
>
> In general routing policy is pretty straight forward with active / backup
> routing policy.
>
>
>
> However when optimal active/ active all active routing optimization is
> required where geographic region based routing is desired and if routing
> control optimization is desired between both peering AS’s , the policy can
> become very complex and unmanageable.  If routing control policy is on one
> side only, the other side AS back hauls based on peering policy the policy
> is complex but manageable.  However optimized routing so both sides of
> peering does not back haul can be very complex.
>
>
>
> There are many peering scenario and this one above  is a dual transit as
> all active peering policy scenario.
>
>
>
> The above relates to AFI /SAFI 1/1 2/1 and can be accomplished with
> standard communities.
>
>
>
> [HC]: The above can be achieved using standard communities, but can be
> very complex as you said. Using the RPD in stead of the standard
> communities can reduce the complexity greatly and can adjust the route
> attributes without configurations for some parts of the network to redirect
> the service traffic from the heavy-loaded links to the links with light
> traffic. Thus the network resource is effectively used and the QoE is
> improved. For example, in one metro area network without using the RPD,
> planning and implementing a redirection of a service traffic (coming to and
> exiting from a router) takes days. After the RPD is deployed in the network
> with a controller NCE, this takes just minutes instead of days.
>
>
>
>
> I would say auto adjustment of active routing paths routing where routing
> control is required on both sides of peering to enable optimized paths
>
>
>
> [HC]: The RPD can be used for automatically adjusting active routing paths
> on the both sides of a peer if it is required and permitted. In the case
> where the two sides of a peer are two different ISPs (say ISPa and ISPb)
> and each side (ISP) does not allow the other to control its network, when
> ISPa adjusts routing paths, it can not control the network owned by ISPb
> directly.
>

   Gyan> Understood.  In this particular case it would have to be
coordinated as exchange is between administrative domains, but in this case
RPD automatic large communities policy update can be accomplished via
centralized PCE controller.

I noticed you have AFI/SAFI codepoint request for RPD.  Would this be just
a SAFI request similar to any other SAFI such as BGP flowspec safi 133 134

https://www.iana.org/assignments/safi-namespace/safi-namespace.xhtml

It would nice to see maybe in an appendix an example of the large
> communities atoms containers of the RPD automated proxy use case of how a
> complex policy translated into RPD policy provides tremendous simplicity in
> provisioning policy updates.
>
>
>
> Kind Regards
>
>
>
> Gyan
>
>
>
>
> <https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.verizon.com%2F&data=02%7C01%7Chuaimo.chen%40futurewei.com%7C07a9442b208a4e85cc3108d81c9b0415%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637290799405571469&sdata=wTNqdsqeXJVkHqK%2FZ0iIDAstepTYRxWtDoxX%2FSOg9HY%3D&reserved=0>
>
>
> <https://www.google.com/maps/search/13101+Columbia+Pike+%0D%0A+Silver+Spring,+MD?entry=gmail&source=g>*Gyan
> Mishra*
>
> *Network Solutions Architect *
>
>
>
> *M 301 502-1347 13101 Columbia Pike
> <https://www.google.com/maps/search/13101+Columbia+Pike+%0D%0A+Silver+Spring,+MD?entry=gmail&source=g>
> *Silver Spring, MD
> <https://www.google.com/maps/search/13101+Columbia+Pike+%0D%0A+Silver+Spring,+MD?entry=gmail&source=g>
>
>
>
-- 

<http://www.verizon.com/>

*Gyan Mishra*

*Network Solutions A**rchitect *



*M 301 502-134713101 Columbia Pike *Silver Spring, MD