Re: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)

["Wanghaibo (Rainsword)" <[e-mail]>]

Hi Gyan,

Thanks for your comments.

 “So now this use case given could use Roberts Large communities draft and apply this RPD  new routing policy AFI SAFI to make changes to a complex IP policy deployment now becomes very simple deploy and manage for day to day operations.”
This is actually a rather precise usage summarization of this draft.

I’d state a little bit more about the rationale behind it.

So in a nutshell, the RPD does more than simply distributing route policy using a new BGP AFI/SAFI, more importantly, it introduces an independent procedure to the route policy process. Conventionally, either by using CLI or Netconf to configure a route policy, the policy change finally is inserted somewhere in the existing policy structure. Finding the right place for the policy to correctly take effect could be truly a hard work, since the policy can be extremely complex as pointed by Gyan. So distributing policy by using RPD jumps out of the box, it adds a final check procedure after the existing policy processing. Thus, it reliefs you from the complex and error-prone policy management, and guarantees you with a deterministic outcome.

Regards,
Haibo

From: Idr [mailto:[e-mail]] On Behalf Of Gyan Mishra
Sent: Saturday, July 25, 2020 6:35 AM
To: Robert Raszuk <[e-mail]>
Cc: Jakob Heitz (jheitz) <jheitz=[e-mail]>; Huaimo Chen <[e-mail]>; Susan Hares <[e-mail]>; [e-mail]
Subject: Re: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)


Authors

I read through the draft again and based on the comments thus far trying to see what gap this draft is addressing that does not exist exist today.

I have commented on this draft earlier on before WGLC and from what I remember the result of the comments was that this draft is for Non SR or MPLS and solely for pure IP based networks that require routing policy push and no traffic steering in the context of PCE based path steering instantiation that already exists today.

This is purely a policy push concept using BGP large communities new path attribute container to accomplish the task.

So with that the idea is that this would accomplished by a centralized controller.

Looking at controller based instantiation you start thinking PCE and BGP LS for IGP path attributes propagation to the PCE, however in this particular case we are not doing any traffic steering.

So from thinking PCE and IP based steering got me thinking about this draft which I see a few authors are on this draft as well.

https://tools.ietf.org/html/draft-ietf-teas-pce-native-ip-09

So bottom line is this new MP Reach AFI SAFI code point assigned for IP based routing control point only which can be managed with a centralized controller doing the policy push via the new AFI SAFI code point.

One complex design I had worked on years ago which I mentioned in the prior thread was based on standard community architecture where both sides are a control point  multi active routing based on communities set.  Imagine you have two MPLS core transports and have a edge CE that is connected as many different locations all around the world and you want you tag all CEs off each core with discrete communities and now you have a special perpending pattern let’s say it’s 3 ties to make it easy and it’s 0x 2x 3x and so now you match communities and set a prepend.  So now you duplicate that on both sides so you don’t backhaul across the other core so you geographically optimally route between each core over all the many tie points.  This does get extremely complex but back then it worked and was all accomplished via standard communities and to this day is still working, however of course management of the policy changes is still a very very complex task.

So now this use case given could use Roberts Large communities draft and apply this RPD  new routing policy AFI SAFI to make changes to a complex IP policy deployment now becomes very simple deploy and manage for day to day operations.

Based on what I have stated I support advancement of the draft with a caveat.  I would like to add my use case to the draft and I think that would help tremendously in explaining the gap and use case being addressed by this draft.


Kind Regards

Gyan

On Fri, Jul 24, 2020 at 5:24 PM Robert Raszuk <[e-mail]<mailto:[e-mail]>> wrote:

In addition to points already mentioned let's examine what the draft actually defines ... not sure if this is just the tip of the iceberg or full set of match and set tools authors envision to have.

Match:

IPv4 prefix,
IPv6 prefix,
AS-PATH regex
Community

Set:

MED,
Arbitrary AS-PATH


- So first don't we see a need to match on Peer's ASN ?
- Don't we need to match on the peer ? (NLRI itself just matches on the ASBR making it p2p)
- How about to match on link bandwidth to the specific peer ?
- The draft talks about outbound traffic engineering from local AS  ... well that is done typically by setting local preference which is not even mentioned in the document
- The defined community sub-TLV merely covers RFC1997. How about Extended or Large Communities ? No use ? No need to match ?

Kind regards,
R.


On Fri, Jul 24, 2020 at 10:28 PM Jakob Heitz (jheitz) <jheitz=[e-mail]<mailto:[e-mail]>> wrote:
Let me try a different way.
Flowspec is designed to install a filter everywhere to mitigate DDoS.
Even if the filter does not install in a few routers, the DDoS is still mitigated.
It does not matter which nodes install it before other nodes.
Therefore, the spray and pray of BGP is ok for Flowspec.

Routing policy distribution could cause unpleasant routing transients if
it is not installed or installed late on random nodes.

Regards,
Jakob.

From: Wanghaibo (Rainsword) <[e-mail]<mailto:[e-mail]>>
Sent: Thursday, July 23, 2020 11:23 PM
To: Jakob Heitz (jheitz) <[e-mail]<mailto:[e-mail]>>; Huaimo Chen <[e-mail]<mailto:[e-mail]>>; Susan Hares <[e-mail]<mailto:[e-mail]>>; idr@ietf..org
Subject: RE: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)

Hi Jakob,

Can you give some more details about "the route and the flowspec spray to the same places“

Regards,
Haibo

From: Jakob Heitz (jheitz) [mailto:[e-mail]]
Sent: Friday, July 24, 2020 1:31 AM
To: Wanghaibo (Rainsword) <[e-mail]<mailto:[e-mail]>>; Huaimo Chen <[e-mail]<mailto:[e-mail]>>; Susan Hares <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: RE: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)

You're missing the point.

The fact that BGP is spray and pray doesn't matter, because the route and the
flowspec spray to the same places.

Regards,
Jakob.

From: Wanghaibo (Rainsword) <[e-mail]<mailto:[e-mail]>>
Sent: Thursday, July 23, 2020 12:02 AM
To: Jakob Heitz (jheitz) <[e-mail]<mailto:[e-mail]>>; Huaimo Chen <[e-mail]<mailto:[e-mail]>>; Susan Hares <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: RE: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)

Hi Jakob,

1.  Flowspec’s validation is used to check whether a device can learn the Flowspec routes from an EBGP peer, but the validation can be performed only for the component of the destination type.
    In practice, the centralized server or controller is often used to send FlowSpec routes to devices.
2.  RPD and SR-Policy also have their own validation. That is, route targets are used to check whether information is sent to the expected node.

Regards,
Haibo

From: Idr [mailto:[e-mail]<mailto:idr-bounces@ietf....org>] On Behalf Of Jakob Heitz (jheitz)
Sent: Tuesday, July 21, 2020 12:52 PM
To: Huaimo Chen <[e-mail]<mailto:[e-mail]>>; Susan Hares <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: Re: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)

There is an important difference between RPD and Flowspec.
https://tools.ietf.org/html/rfc5575#section-6
states:
   A flow specification NLRI must be validated such that it is
   considered feasible if and only if:

   a) The originator of the flow specification matches the originator of
      the best-match unicast route for the destination prefix embedded
      in the flow specification.

   b) There are no more specific unicast routes, when compared with the
      flow destination prefix, that have been received from a different
      neighboring AS than the best-match unicast route, which has been
      determined in step a).

Effectively, the advertisement of the route takes the same vector as the
advertisement of the matching flowspec. Therefore, if the flowspec did not
reach a node, then the route likely didn't either, so it doesn't matter.

The fact that BGP is spray and pray doesn't matter, because the route and the
flowspec spray to the same places.

RPD policy distribution has no such validation rule.

SR policy distribution suffers from the same problem.


Regards,
Jakob.

From: Huaimo Chen <[e-mail]<mailto:[e-mail]>>
Sent: Monday, July 20, 2020 9:01 PM
To: Jakob Heitz (jheitz) <[e-mail]<mailto:[e-mail]>>; Susan Hares <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: Re: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)

Hi Jakob,

    Thank you very much for your valuable comments.
    Our answers/explanations are inline below with prefix [HC].

Best Regards,
Huaimo on behalf of co-authors
________________________________
From: Idr <[e-mail]<mailto:[e-mail]>> on behalf of Jakob Heitz (jheitz) <jheitz=[e-mail]..org<mailto:jheitz=40cisco.com@dmarc..ietf.org>>
Sent: Thursday, July 16, 2020 9:01 PM
To: Susan Hares <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)


BGP seems the wrong way to distribute routing policy.



[HC]: It seems that BGP flow spec has been used widely to distribute policies for redirecting the traffic. It seems work well without some mechanisms in Netconf. BGP RPD should be similar to BGP flow spec.  BGP SR Policy is on the same train.



IETF has already defined a way to distribute configuration: Netconf.

Netconf provides needed features that BGP does not have:

- Atomic Transactions:

  If one configuration item fails, they all fail.

  They all either succeed or all fail. There is no partial success.

  Multiple configurations in one transaction are applied at the same time.

   . This avoids non-deterministic transient behavior between application of the first policy and the last.

- Feedback:

  BGP is "spray and pray".

  Netconf provides an acknowledgement that the config either failed or was applied,

  which then allows the controller to take the next steps with

  reliable information about what configuration exists in the network.

- Persistence:

  If the BGP session were to go down, all the configuration it sent will be implicitly withdrawn.



If another AS would not allow a foreign AS to configure it with netconf,

it would not allow it with RPD either.



There are already ways in BGP for an AS to signal preference across AS boundaries:

Med, AS-path length, communities.



[HC]: Netconf can be used to distribute configurations from a controller to the devices in a network. BGP RPD as an alternative option, may have some advantages in some cases. For example, in the case where BGP as a controller, BGP RPD seems more suitable. Using BGP RPD to control/redirect the traffic dynamically in real time may be more effective.



Regards,

Jakob.



From: Idr <[e-mail]<mailto:[e-mail]>> On Behalf Of Susan Hares
Sent: Wednesday, July 15, 2020 6:11 AM
To: [e-mail]<mailto:[e-mail]>
Subject: [Idr] WG LC on draft-ietf-idr-rpd-05.txt (7/15 to 7/29/2020)



This begins a 2 week WG LC on draft-ietf-idr-rpd

from 7/15 to 7/29/2020.  You can obtain this draft at:

https://datatracker.ietf.org/doc/draft-ietf-idr-rpd/<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fdraft-ietf-idr-rpd%2F&data=02%7C01%7Chuaimo.chen%40futurewei.com%7C12cf72daefe0446d5a7908d829ed0a36%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637305445341383523&sdata=3LvgG6xwElOv27jGetqpyk8ftRub%2B%2B4Ui31Yt8wN87A%3D&reserved=0>



This draft defines a new AFI/SAFI and new atoms

for the Wide Communities.  This WG LC has been delayed

as I waited for a resubmission of the Wide Communities draft.

I had hoped to do these 2 WG LC in parallel.



I’ve not received the Wide Communities draft, but we will

start this WGLC to provide feedback to the authors.

We may have to run a short follow-up to this WG LC

If there are changes to the Wide Communities draft during

Its WG LC.



There is an IPR statement on this draft.



In your responses please answer the following questions:



1) Do you feel this draft has an solution that is acceptable

   With the IPR as a WG RFC?



2) Do you feel this draft is ready to publish?



3) Do you know of implementations of this draft?



4) Do you know of deployments of this draft?

If so, is this feature useful in the deploy ments.



5) Do you feel that Wide Communities is ready for

Publication?



Cheerily, Susan Hares
_______________________________________________
Idr mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/idr
_______________________________________________
Idr mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/idr
--

[图像已被发件人删除。]<http://www.verizon.com/>

Gyan Mishra

Network Solutions Architect

M 301 502-1347
13101 Columbia Pike
Silver Spring, MD