Re: [spring] 答复: CRH is back to the SPRING Use-Case - Re: Size of CR in CRH

Aijun Wang <wangaijun@tsinghua.org.cn> Sun, 24 May 2020 08:17 UTC

Return-Path: <wangaijun@tsinghua.org.cn>
X-Original-To: ipv6@ietfa.amsl.com
Delivered-To: ipv6@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id E27A43A07AC; Sun, 24 May 2020 01:17:26 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.895
X-Spam-Level:
X-Spam-Status: No, score=-1.895 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, HTML_MESSAGE=0.001, MIME_QP_LONG_LINE=0.001, RCVD_IN_MSPIKE_H4=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=unavailable autolearn_force=no
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id FyRMXOvPZ_ST; Sun, 24 May 2020 01:17:21 -0700 (PDT)
Received: from m176115.mail.qiye.163.com (m176115.mail.qiye.163.com [59.111.176.115]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 415113A073D; Sun, 24 May 2020 01:17:19 -0700 (PDT)
Received: from [240.0.0.1] (unknown [45.11.3.222]) by m176115.mail.qiye.163.com (Hmail) with ESMTPA id B3F42664576; Sun, 24 May 2020 16:17:06 +0800 (CST)
Content-Type: multipart/alternative; boundary=Apple-Mail-90B8529F-91A3-49CF-B860-FA10402F39B5
Content-Transfer-Encoding: 7bit
From: Aijun Wang <wangaijun@tsinghua.org.cn>
Mime-Version: 1.0 (1.0)
Subject: =?utf-8?Q?Re:_[spring]__=E7=AD=94=E5=A4=8D:__CRH_is_back_to_the_?= =?utf-8?Q?SPRING_Use-Case_-_Re:_Size_of_CR_in_CRH?=
Date: Sun, 24 May 2020 16:16:50 +0800
Message-Id: <30057A77-075A-4E53-8055-6A6D9502823A@tsinghua.org.cn>
References: AB4B404F-BE99-4053-8A5A-01EC90C9B0D5
Cc: Andrew Alston <Andrew.Alston@liquidtelecom.com>, "Ketan Talaulikar (ketant)" <ketant=40cisco.com@dmarc.ietf.org>, "Joel M. Halpern" <jmh@joelhalpern.com>, rtg-ads <rtg-ads@ietf.org>, spring <spring@ietf.org>, 6man <6man@ietf.org>
In-Reply-To: AB4B404F-BE99-4053-8A5A-01EC90C9B0D5
To: "Pengshuping (Peng Shuping)" <pengshuping@huawei.com>
X-Mailer: iPhone Mail (17E262)
X-HM-Spam-Status: e1kfGhgUHx5ZQUtXWQgYFAkeWUFZVkpVQ0xDS0tKSUpIS0xITEhZV1koWU FKTEtLSjdXWS1ZQUlXWQkOFx4IWUFZNTQpNjo3JCkuNz5ZBg++
X-HM-Sender-Digest: e1kMHhlZQR0aFwgeV1kSHx4VD1lBWUc6Phw6Syo*Lzg4Ti48IUo2I00r E0JPChRVSlVKTkJLSEtDSUhOSkhOVTMWGhIXVQwaFRwaEhEOFTsPCBIVHBMOGlUUCRxVGBVFWVdZ EgtZQVlPTlVKSlVIVUlJSVlXWQgBWUFOS0lITzcG
X-HM-Tid: 0a7245c1c6b59373kuwsb3f42664576
Archived-At: <https://mailarchive.ietf.org/arch/msg/ipv6/bgQYRWjdLQAeAB7DqC47ml0XH0M>
X-BeenThere: ipv6@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: "IPv6 Maintenance Working Group \(6man\)" <ipv6.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ipv6>, <mailto:ipv6-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ipv6/>
List-Post: <mailto:ipv6@ietf.org>
List-Help: <mailto:ipv6-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ipv6>, <mailto:ipv6-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 24 May 2020 08:17:31 -0000

Hi, 
We know the advantages of these proposals. But we also pay more attention to their deployment in large network.
Considering the flows that needs to be steered away the default path are often small packets, adding the path list overhead on it is not efficient. And we need to add these overhead to every packet of the flow...

I am wondering why we go from one extreme solution(keep state on network device) to another extreme solution(keep state on every packet)? Network device has the capability to keep state, why put these capabilities aside?

There are other aspects that we needs to consider, for example the SID allocation across the AS boundary, The Path MTU, the Binding SID, the complexities of different behavior based on these SIDS. All these Issues have been discussed intensely in the mail list but it seems one solution will induce another issue and formed one endless loop... ...

My thought is that if we depends on the central control/PCE  to accomplish the E2E QoS assurance, we should simplify the protocol, not just the number of them, but also the complexity of implementation and deployment.

Aijun Wang
China Telecom

> On May 23, 2020, at 08:23, Pengshuping (Peng Shuping) <pengshuping@huawei.com> wrote:
> 
> 
> Hi,
> 
> Please see inline.
> 
> --------------------------------------------------
> 彭书萍 Peng Shuping
> Mobile: +86-18210364128(优先)
> Email: pengshuping@huawei.com
> 
> 发件人:Aijun Wang <wangaijun@tsinghua.org.cn>
> 收件人:'Andrew Alston' <Andrew.Alston@liquidtelecom.com>;'Ketan Talaulikar (ketant)' <ketant=40cisco.com@dmarc.ietf.org>;'Joel M. Halpern' <jmh@joelhalpern.com>
> 抄 送:rtg-ads <rtg-ads@ietf.org>;spring <spring@ietf.org>;'6man' <6man@ietf.org>
> 时 间:2020-05-22 23:55:46
> 主 题:[spring] 答复: CRH is back to the SPRING Use-Case - Re: Size of CR in CRH
> 
> Hi, Andrew and all:
> 
> Is using the “Routing Header” to maneuver the path of packet a good idea?
> 
> From the viewpoint of the service provider, if we depend on the central controller/PCE to assure the QoS of application, we should simplify the network design and the network device.
> 
> Currently, both SRH and CRH deployment requires the presence of SDN controller/PCE, but the network/device complexity is not decreased. How operator can benefit from the adoption of these technologies?
> 
> 
> 
> [Shuping] Protocols are simplified. Only need to care about IGP and BGP. 
> 
> 
> 
> You have also mentioned the cloud provider. What their requirements is the openness/standard API from the device, this can give the most flexibility for network programming.
> 
> Why we don’t dive into this direction? There are several NorthBound interfaces/protocols exist in IETF standard.
> 
> 
> 
> [Shuping] PCEP and BGP can be taken as the standard API of an SRv6 network, by which you can program your network. 
> 
> 
> 
> For CRH and SRH, I think there are some similarity between them. CRH does not conflict with the IPv6 address, but requires the map table exist. Many experts will dispute again the similar problem.
> 
> Is CRH the brick that cloud provider want?
> 
> 
> 
> [Shuping] SRv6 into DC is also appealing since it enables end-to-end true network and cloud convergence. 
> 
> 
> 
> Best Regards.
> 
> Aijun Wang
> 
> China Telecom
> 
> 发件人: spring-bounces@ietf.org [mailto:spring-bounces@ietf.org] 代表 Andrew Alston
> 发送时间: 2020年5月22日 17:25
> 收件人: Andrew Alston; Ketan Talaulikar (ketant); Joel M. Halpern
> 抄送: rtg-ads@ietf.org; spring@ietf.org; 6man
> 主题: Re: [spring] CRH is back to the SPRING Use-Case - Re: Size of CR in CRH
> 
> Just as a followup – which I think illustrates my point here.
> 
> https://blogs.cisco.com/sp/cisco-goes-sonic-on-new-networking-platforms
> 
> If operators wanted fait accompli – why is it that Cisco – and other vendors – are starting to provide ASIC based API’s such that cloud providers and operators can do what they need on devices – rather than the dictates of vendors? I think the very fact that the above article exists – illustrates the desire for simple building blocks.
> 
> Thanks
> 
> Andrew
> 
> From: spring <spring-bounces@ietf.org> On Behalf Of Andrew Alston
> Sent: Friday, 22 May 2020 12:15
> To: Ketan Talaulikar (ketant) <ketant=40cisco.com@dmarc.ietf.org>rg>; Joel M. Halpern <jmh@joelhalpern.com>
> Cc: rtg-ads@ietf.org; spring@ietf.org; 6man <6man@ietf.org>
> Subject: Re: [spring] CRH is back to the SPRING Use-Case - Re: Size of CR in CRH
> 
> Actually Ketan,
> 
> As an operator – I am looking for the tyre – I want the building block – because it allows me to both use what the vendors build on top of it – and to build my own stuff on top of it that is specific to my needs.
> 
> The tyre association is one such association – the brick is another – the brick was invented long before anyone created an architectural diagram to build a building. CRH is a brick – that can be shaped to many purposes.
> 
> And as an operator – it is *exactly* what I want – without the dictates of what any vendor says I can do with it.
> 
> There are unique conditions in every operator environment – and if an operator is given the right bricks – and those bricks can be used to build something that is inter-operable across the vendors and within their specific domain – it is that building that gives the operator their competitive advantage. If however, all operators are handed a completed building – its much harder to build something that differentiates.
> 
> I must point out – in a world where we are seeing more and more white box technology – and where content providers are moving to building their own things pretty quickly – why do you think that is? One reason I believe is because the requirements are not the same everywhere – and I have long believed that a standard should be such that an operator can take that standard – and use it within his specific domain to build on top of – and rather we have a standard brick such that the solutions that are built are portable – than every operator inventing their own bricks and nothing working together.
> 
> Again – the reason I support CRH – is because of the flexibility and simplicity it affords me as an operator – and because – we have specific use cases and things we wish to build on top of it – some of which – are pretty standard and happily in the public domain – other things – which are specific internally. And I for one right now – am far more willing as an operator to pay for the bricks that let me build something that is competitive in the market than pay for fait accompli that is nothing more than the dictates for what vendors think I want.
> 
> And yes – I support the divorce of CRH for all things SPRING related – because the use cases beyond segment routing are multiple – and I have absolutely zero desire – plan – or anything else to use something like the PGM draft on my network and impose that level of overhead, complexity, overloading of the v6 address and what I view as a general corruption of the v6 architecture. What I want – is a simple routing header – that I can build on – and that’s the end of the architecture that I want dictated to me.
> 
> Thanks
> 
> Andrew
> 
> From: ipv6 <ipv6-bounces@ietf.org> On Behalf Of Ketan Talaulikar (ketant)
> Sent: Friday, 22 May 2020 08:24
> To: Joel M. Halpern <jmh@joelhalpern.com>
> Cc: rtg-ads@ietf.org; spring@ietf.org; 6man <6man@ietf.org>
> Subject: RE: [spring] CRH is back to the SPRING Use-Case - Re: Size of CR in CRH
> 
> I am thinking that the operators would be looking for the car and not the tyre?
> 
> Thanks,
> Ketan
> 
> -----Original Message-----
> From: Ketan Talaulikar (ketant) 
> Sent: 22 May 2020 10:55
> To: 'Joel M. Halpern' <jmh@joelhalpern.com>
> Cc: spring@ietf.org; 6man <6man@ietf.org>rg>; rtg-ads@ietf.org
> Subject: RE: [spring] CRH is back to the SPRING Use-Case - Re: Size of CR in CRH
> 
> Hi Joel,
> 
> I'll point you to RFC7855, RFC8355 and RFC8402 that cover both the data-planes for Spring. Then the RFC8354 which is focussed on SRv6. All this body of work along with a whole lot of discussion and brainstorming happening in the Spring WG provided the architecture, use-cases, applicability and requirements for SRH (RFC8754).
> 
> It may be so that many people in 6man focussed on only the IPv6 specific aspects as is their design expertise. But there were others (in 6man, Spring and other WGs) that were able to look at the solution in a holistic manner thanks to the body of work behind it.
> 
> Net-PGM builds on top of RFC8402 and RFC8754.
> 
> To give a real world analogy, let us understand what kind of a car we are trying to build (to carry goods/passengers or both and how much/many, what terrain it is meant for, what weather/environment conditions, how much speed/performance/fuel efficiency parameters required, etc.) before we start designing tyres for it.
> 
> Thanks,
> Ketan 
> 
> -----Original Message-----
> From: Joel M. Halpern <jmh@joelhalpern.com>
> Sent: 22 May 2020 10:02
> To: Ketan Talaulikar (ketant) <ketant@cisco.com>
> Cc: spring@ietf.org; 6man <6man@ietf.org>rg>; rtg-ads@ietf.org
> Subject: Re: [spring] CRH is back to the SPRING Use-Case - Re: Size of CR in CRH
> 
> Ketan, I am trying to figure out which documents you think were adopted and approved elsewhere to drive the 6man work on SRH.
> 
> I did find RFC 8354, which was a use case. It is not a problem statement. It is most definitely not an architecture. The only architecture documents I can find are general SR documents. Those did not justify a need for SRH. And I (at least) did not object to SRH on the basis of that gap.
> 
> Yes, SRH normatively references 8402. But 8402 does not drive any need for SRH. In fact, the actual text references to SRH are fairly cursory. 
> (The most significant is some terminology.)
> 
> In fact, as far as I can tell, the ties are such that there is no evidence in the documents that SPRING had any say in SRH. (the reality is more complex, I grant you. But there was no formal approval or signoff.)
> 
> As far as I can tell, there was no formal approval of anything by SPRING that can be read as a request to 6man to work on SRH. (Do remember that the SRH document was adopted by 6man in December of 2015.) The network programming draft did not even appear at 00 until March of 2017, 15 months later.
> 
> How, given this history, can you claim that CRH needs something more. 
> We have operators asking for this.
> 
> Yours,
> Joel
> 
> On 5/21/2020 11:53 PM, Ketan Talaulikar (ketant) wrote:
> > Hi Bob,
> > 
> > Perhaps I will try to make my case to you (and everyone else here) … 
> > one last time.
> > 
> > This is how I've seen RH work being done in 6man until now (in a 
> > matter that fits its charter).
> > 
> > 1) There is a WG (not 6man) that defines the problem statement, 
> > use-cases and architecture that requires RH
> > 
> > 2) The 6man being the experts on IPv6 design, either take up the 
> > document that specifies that RH (or even if it is done in another WG, 
> > reviews it).
> > 
> > So 6man has always had work done in (1) to reference and lean upon 
> > when doing (2).
> > 
> > My argument of the shortcut in the case of this specific adoption is 
> > that we don't have (1).
> > 
> > It is not in 6man charter nor expertise to take up (1) because CRH is 
> > not purely IPv6 work. It is not meant for "Internet" but a specific 
> > "limited domain". The SIDs that it introduces is a new "mapping ID"
> > concept. It is not an IPv6 address and neither it is MPLS. This is a
> > *_Routing_* Header and part of a new Source *_Routing_* solution.
> > 
> > Therefore, without (1) being made available to 6man, I believe that 
> > working on (2) in 6man is to me a shortcutting of the IETF technical 
> > review process (specifically of the *_Routing_* area in this case) for 
> > a solution and does not provide the necessary reference for 6man to work on.
> > 
> > Why the rush?
> > 
> > I close my arguments.
> > 
> > Thanks,
> > 
> > Ketan
> > 
> > -----Original Message-----
> > From: Bob Hinden <bob.hinden@gmail.com>
> > Sent: 22 May 2020 09:03
> > To: Ketan Talaulikar (ketant) <ketant@cisco.com>
> > Cc: Bob Hinden <bob.hinden@gmail.com>om>; Brian Carpenter 
> > <brian.e.carpenter@gmail.com>om>; Ron Bonica <rbonica@juniper.net>et>; 
> > Chengli (Cheng Li) <c.l@huawei.com>om>; Zafar Ali (zali) 
> > <zali@cisco.com>om>; Robert Raszuk <robert@raszuk.net>et>; spring@ietf.org; 
> > 6man <6man@ietf.org>
> > Subject: Re: [spring] CRH is back to the SPRING Use-Case - Re: Size of 
> > CR in CRH
> > 
> > Ketan,
> > 
> > > On May 21, 2020, at 8:12 PM, Ketan Talaulikar (ketant) 
> > <ketant=40cisco.com@dmarc.ietf.org
> > <mailto:ketant=40cisco.com@dmarc.ietf.org>> wrote:
> > 
> > >
> > 
> > > Hi Brian,
> > 
> > >
> > 
> > > Please see my previous response to your comments.
> > 
> > >
> > 
> > > My argument is not legalistic. I am not as experience in IETF work 
> > as you and Bob are. But what I understand is that the reason why we 
> > have these "legal" process of charters and BoF is to enable a proper 
> > technical discussion with the right context and details of the 
> > proposal presented for review of the community.
> > 
> > >
> > 
> > > I do not see how shortcutting them helps anyone and I wonder why it 
> > is being done in this case?
> > 
> > There is no short cutting here. The adoption call is to determine if 
> > there is interest in the w.g. to take this work into 6man. If it 
> > becomes a w.g. draft, then the w.g. is responsible to decide what 
> > happens next.
> > 
> > It’s a first step, it is not a decision to publish it.
> > 
> > Bob (w/ w.g. chair hat on)
> > 
> > >
> > 
> > > Thanks,
> > 
> > > Ketan
> > 
> > >
> > 
> > > -----Original Message-----
> > 
> > > From: Brian E Carpenter <brian.e.carpenter@gmail..com 
> > <mailto:brian.e.carpenter@gmail.com>>
> > 
> > > Sent: 22 May 2020 04:18
> > 
> > > To: Ketan Talaulikar (ketant) <ketant@cisco.com 
> > <mailto:ketant@cisco.com>>; Ron Bonica <rbonica@juniper.net 
> > <mailto:rbonica@juniper.net>>; Chengli (Cheng Li) <c.l@huawei.com 
> > <mailto:c.l@huawei.com>>; Zafar Ali (zali) <zali@cisco.com 
> > <mailto:zali@cisco.com>>; Robert Raszuk <robert@raszuk.net 
> > <mailto:robert@raszuk.net>>
> > 
> > > Cc: spring@ietf.org <mailto:spring@ietf.org>; 6man <6man@ietf.org 
> > <mailto:6man@ietf.org>>
> > 
> > > Subject: Re: CRH is back to the SPRING Use-Case - Re: Size of CR in 
> > CRH
> > 
> > >
> > 
> > > On 22-May-20 05:26, Ketan Talaulikar (ketant) wrote:
> > 
> > > ....> It is the 6man charter that precludes it from defining a new 
> > Source Routing solution..
> > 
> > >> “It is not chartered to develop major changes or additions to the
> > IPv6 specifications.”
> > 
> > >
> > 
> > > If this addition was major, that would be true. But adding a new RH 
> > type is well within the scope of maintenance, IMHO. We have already 
> > done it quite recently.
> > 
> > >
> > 
> > > In any case, legalistic arguments about WG charters are really not 
> > how we should take technical decisions.
> > 
> > >
> > 
> > > Regards
> > 
> > >   Brian
> > 
> > >
> > 
> > >
> > 
> > > _______________________________________________
> > 
> > > spring mailing list
> > 
> > > spring@ietf.org <mailto:spring@ietf.org>
> > 
> > > https://www.ietf.org/mailman/listinfo/spring
> > 
> > 
> > --------------------------------------------------------------------
> > IETF IPv6 working group mailing list
> > ipv6@ietf.org
> > Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> > --------------------------------------------------------------------
> > 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> ipv6@ietf.org
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
> 
> 
> _______________________________________________
> spring mailing list
> spring@ietf.org
> https://www.ietf.org/mailman/listinfo/spring