Re: [mpls] [spring] [SPRING] Query related to SR Architecture

"Joel M. Halpern" <jmh@joelhalpern.com> Thu, 10 September 2015 17:11 UTC

Return-Path: <jmh@joelhalpern.com>
X-Original-To: mpls@ietfa.amsl.com
Delivered-To: mpls@ietfa.amsl.com
Received: from localhost (ietfa.amsl.com [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id A96EE1B30A8; Thu, 10 Sep 2015 10:11:20 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.602
X-Spam-Level:
X-Spam-Status: No, score=-2.602 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, RCVD_IN_DNSWL_LOW=-0.7, SPF_HELO_PASS=-0.001, SPF_PASS=-0.001] autolearn=ham
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 yMuqgKYKwCIG; Thu, 10 Sep 2015 10:11:17 -0700 (PDT)
Received: from mailb2.tigertech.net (mailb2.tigertech.net [208.80.4.154]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 76E451B5603; Thu, 10 Sep 2015 10:11:17 -0700 (PDT)
Received: from localhost (localhost [127.0.0.1]) by mailb2.tigertech.net (Postfix) with ESMTP id 528F67800E0; Thu, 10 Sep 2015 10:11:17 -0700 (PDT)
X-Virus-Scanned: Debian amavisd-new at b2.tigertech.net
Received: from Joels-MacBook-Pro.local (209-255-163-147.ip.mcleodusa.net [209.255.163.147]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by mailb2.tigertech.net (Postfix) with ESMTPSA id 7A0737800DF; Thu, 10 Sep 2015 10:11:16 -0700 (PDT)
To: "Anil Kumar S N (VRP Network BL)" <anil.sn@huawei.com>, Gaurav agrawal <gaurav.agrawal@huawei.com>, Alexander Vainshtein <Alexander.Vainshtein@ecitele.com>
References: <544F5E3F-82AD-49BA-A83B-201DE49A08A6@juniper.net> <327562D94EA7BF428CD805F338C31EF06C0496F5@nkgeml512-mbx.china.huawei.com> <99EAE216-DB6C-4AFD-8E5C-E834D68CBF52@juniper.net> <327562D94EA7BF428CD805F338C31EF06C04975F@nkgeml512-mbx.china.huawei.com> <151A634F-9F72-40E6-AAC7-94F66F2CDFF5@juniper.net> <D2171FC6.2DF9A%acee@cisco.com> <327562D94EA7BF428CD805F338C31EF06C049BD1@nkgeml512-mbx.china.huawei.com>
From: "Joel M. Halpern" <jmh@joelhalpern.com>
Message-ID: <55F1B9B2.3030102@joelhalpern.com>
Date: Thu, 10 Sep 2015 13:11:14 -0400
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:38.0) Gecko/20100101 Thunderbird/38.2.0
MIME-Version: 1.0
In-Reply-To: <327562D94EA7BF428CD805F338C31EF06C049BD1@nkgeml512-mbx.china.huawei.com>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: <http://mailarchive.ietf.org/arch/msg/mpls/vKtQAvr_vk-s7zApiuqCXTTJh-w>
Cc: "mpls@ietf.org" <mpls@ietf.org>, "spring@ietf.org" <spring@ietf.org>
Subject: Re: [mpls] [spring] [SPRING] Query related to SR Architecture
X-BeenThere: mpls@ietf.org
X-Mailman-Version: 2.1.15
Precedence: list
List-Id: Multi-Protocol Label Switching WG <mpls.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/mpls>, <mailto:mpls-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/mpls/>
List-Post: <mailto:mpls@ietf.org>
List-Help: <mailto:mpls-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/mpls>, <mailto:mpls-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 10 Sep 2015 17:11:20 -0000

Anil,
     The task of using MPLS for carrying service chaining can be 
addressed in multiple ways without making odd assumptions or changing 
MPLS semantics.
     First, please, do not assume that "all nodes deliver the same 
service".  that is not the model being used by IETF SFC nor by ETSI NFV.
     Second, there are at least two different ways to use MPLS as an SFC 
transport across multiple service functions without violating the 
existing MPLS architecture.  One could use MPLS Spring, for example 
using the proposal from Xuxiaohu as a trqnsport.  One could also use an 
MPLS LSP that traverses several Service function forwarders and service 
functions.  This could be done either with MPLS support on the Service 
Function, or with service functions which simply preserve the MPLS 
header, and the service function forwarders take care of the remapping 
that is part of MPLS.  Either of these preserve the MPLS architecture 
that the local node forwards only based on the top label.

     If one wants to preserve an MPLS label that is external to the 
service chaining system, that is probably best done via interaction with 
the ingress and egress systems.  If one is certain that the entire 
domain supports MPLS, one could keep the external stack, but have 
special processing at the ingress and egress of the SFC domain for 
adding and removing the NSH header, but personally I would not recommend 
that approach.

Yours,
Joel

On 9/10/15 1:04 PM, Anil Kumar S N (VRP Network BL) wrote:
> Hi Acee,
>
>      Thanks for your inputs. I might be wrong, correct me if i am.
>
>          Each node can advertise two different types service label for a
> same service.
>          a)  Service label after poping it, perform the service
> continue  with forwarding (current case)
>          b)  New kind of service label after poping it, perform/schedule
> the service and before forwarding push back the neighbors local service
> label.
>
> Will this have backward compatibility issue ? as we are attaching more
> meaning to a service label which is new.
> (new kind of service label processing will be done only be capable nodes
> who understand this new label type, capability must be exchanged to
> support backward compatibility)
>
> Implementation implications : I could foresee one difficulty but need
> communities opinion
>                                                          The top label
> which is a special service label will be swapped with peer's service
> label but only after forwarding decision is made.
>
> Partial development : This i am not sure at this point of time, i will
> definitely check with with our product team and get back to you ASAP.
>
> With Regards
> Anil S N
> ------------------------------------------------------------------------
> *From:* spring [spring-bounces@ietf.org] on behalf of Acee Lindem (acee)
> [acee@cisco.com]
> *Sent:* Thursday, September 10, 2015 9:46 PM
> *To:* Pushpasis Sarkar; Anil Kumar S N (VRP Network BL); Gaurav agrawal;
> Alexander Vainshtein
> *Cc:* mpls@ietf.org; spring@ietf.org; Vinod Kumar S 70786
> *Subject:* Re: [spring] [SPRING] Query related to SR Architecture
>
> Hi Pushpasis, Anil,
>
> From: spring <spring-bounces@ietf.org <mailto:spring-bounces@ietf.org>>
> on behalf of Pushpasis Sarkar <psarkar@juniper.net
> <mailto:psarkar@juniper.net>>
> Date: Thursday, September 10, 2015 at 11:42 AM
> To: "Anil Kumar S N (VRP Network BL)" <anil.sn@huawei.com
> <mailto:anil.sn@huawei.com>>, Gaurav agrawal <gaurav.agrawal@huawei.com
> <mailto:gaurav.agrawal@huawei.com>>, Alexander Vainshtein
> <Alexander.Vainshtein@ecitele.com <mailto:Alexander.Vainshtein@ecitele.com>>
> Cc: "mpls@ietf.org <mailto:mpls@ietf.org>" <mpls@ietf.org
> <mailto:mpls@ietf.org>>, "spring@ietf.org <mailto:spring@ietf.org>"
> <spring@ietf.org <mailto:spring@ietf.org>>, Vinod Kumar S 70786
> <v70786@notesmail.huawei.com.cn <mailto:v70786@notesmail.huawei.com.cn>>
> Subject: Re: [spring] [SPRING] Query related to SR Architecture
>
>     Hi Anil,
>
>     Inline again
>
>     From: "Anil Kumar S N (VRP Network BL)"
>     Date: Thursday, September 10, 2015 at 8:56 PM
>     To: Pushpasis Sarkar, Gaurav agrawal, Alexander Vainshtein
>     Cc: "mpls@ietf.org <mailto:mpls@ietf.org>", "spring@ietf.org
>     <mailto:spring@ietf.org>", Vinod Kumar S 70786
>     Subject: RE: [spring] [SPRING] Query related to SR Architecture
>
>     Pushpasis,
>
>                      Thank you again, Please see inline */[Anil >>]/*.
>
>     Thanks & Regards
>
>     Anil S N
>
>     “Be liberal in what you accept, and conservative in what you send” -
>     Jon Postel
>
>     *From:*spring [mailto:spring-bounces@ietf.org] *On Behalf Of
>     *Pushpasis Sarkar
>     *Sent:* 10 September 2015 20:15
>     *To:* Anil Kumar S N (VRP Network BL); Gaurav agrawal; Alexander
>     Vainshtein
>     *Cc:* mpls@ietf.org <mailto:mpls@ietf.org>; spring@ietf.org
>     <mailto:spring@ietf.org>; Vinod Kumar S 70786
>     *Subject:* Re: [spring] [SPRING] Query related to SR Architecture
>
>     Hi Anil,
>
>     *From: *"Anil Kumar S N (VRP Network BL)"
>     *Date: *Thursday, September 10, 2015 at 6:17 PM
>     *To: *Pushpasis Sarkar, Gaurav agrawal, Alexander Vainshtein
>     *Cc: *"spring@ietf.org <mailto:spring@ietf.org>", "mpls@ietf.org
>     <mailto:mpls@ietf.org>", Vinod Kumar S 70786
>     *Subject: *RE: [spring] [SPRING] Query related to SR Architecture
>
>     Hi Pushpasis,
>
>                      Thanks a lot for replying.
>
>                      The requirement is still under research stage once
>     it is in a presentable format  will share the details.
>
>     we need every node on the path to perform the same service, we could
>     even define a service globally and allocate one label to it which is
>     understandable by all.
>
>     [Pushpasis] First, it does not make sense to me why all the nodes on
>     the path will execute the same service on the payload? It would make
>     more sense that they do different services on different node. Can
>     you illustrate with an example?  Second, If they all are providing
>     the same service, why need a separate service label?
>
>     */[Anil >>] /**What did you meant by **“**Second, If they all are
>     providing the same service, why need a separate service label?
>     **”**.  What we meant is the case where 10 different services need
>     to be performed by every hop based on based on received service
>     label in the SR packet.*
>
>     [PS2] Ok. So what you are saying is… Each node is capable of N
>     number services. The service label is to only indicate which one it
>     is. Right? It is still not clear why the same service needs to be
>     executed on each transit node.
>
>     **
>
>                      Technically it must be possible right, as top label
>     specifies to service once that service is performed then refer
>     ILM/NHLFE table for forwarding and if service label says before
>     sending out the packet push the label back on the stack.
>
>     [Pushpasis] In MPLS architecture, labels are always of local
>     significance. So the global service label you are talking about MUST
>     be actually a label allocated by the first hop node who has
>     allocated the specific label for the service. Even the Node Segment
>     Label the ingress will use to push the packet is allocated by the
>     first hop node.
>
>     */[Anil >>]  pushing local service label originated by immediate
>     nexthop before transmitting will serve the purpose , each node will
>     push new local service label based on the nexthop as a last step
>     after processing service label & node label./*
>
>     */we want service label on the top of the stack./*
>
>     [PS2] You mean ingress will put the service label and node segment
>     label and send it to first hop. Then first hop will pop the service
>     label, execute the service, look at the next node label and then add
>     the same service label along with node segment label advertised by
>     the second hop for the final destination. And so on. Right? This
>     sounds to convoluted to me.
>
>
> Clearly convoluted. Even if the desired behavior is applicable to every
> node, this represents a unique MPLS forwarding paradigm. It is similar
> to stitching LSPs but not identical.  Before introducing something like
> this, the MPLS WG would have to consider implementation implications,
> partial deployment, and backward compatibility versus the benefits of
> the limited use cases.
>
> Thanks,
> Acee
>
>
>
>
>
>     *//*
>
>                      The Egress device will finally pop both service and
>     node label.
>
>     Hope you are refering to with respect to SFC
>     https://tools.ietf.org/html/draft-xu-spring-sfc-use-case-02”
>
>                      As per below section a node label and service lable
>     combination has to be pushed for each hop if the service is intended
>     to  be performed by each node on the path to destination.
>
>                      If there is any metadata involved for service NSH
>     has to be piggybacked in the packet. SFC don’t solve in reducing
>     number of labels involved.
>
>     [Pushpasis] I agree. Perhaps we need to find a different solution.
>
>     */[Anil >>]  By having service label at the top will solve label
>     stack issues in some cases./*
>
>     [PS2] I don’t think so. Once again I don’t get why all transit nodes
>     need to execute the same service again and again on the same
>     payload? Have you considered the case where 3 (say) different
>     service needs to be applied at the three different service nodes and
>     then finally forwarded to destination? How will you specify three
>     different services without three different service labels?
>
>     Thanks
>     -Pushpasis
>
>     *//*
>
>     *//*
>
>     */would like to know technical issues involved in having service
>     label as top label, Please help us./*
>
>     Thanks
>
>     -Pushpasis
>
>              The service classifier therefore would attach a
>
>              segment list {SID(SN1), SID(SF1), SID(SN2), SID(SF2)} to
>     the packet.
>
>              This segment list is actually represented by a MPLS label
>     stack.  In
>
>              addition, the service classifier could optionally impose
>     metadata on
>
>              the packet through the Network Service Header (NSH)
>
>     Thanks & Regards
>
>     Anil S N
>
>     “Be liberal in what you accept, and conservative in what you send” -
>     Jon Postel
>
>     *From:*Pushpasis Sarkar [mailto:psarkar@juniper.net]
>     *Sent:* 10 September 2015 00:45
>     *To:* Gaurav agrawal; Alexander Vainshtein
>     *Cc:* Anil Kumar S N (VRP Network BL); spring@ietf.org
>     <mailto:spring@ietf.org>; mpls@ietf.org <mailto:mpls@ietf.org>;
>     Vinod Kumar S 70786
>     *Subject:* Re: [spring] [SPRING] Query related to SR Architecture
>
>     Hi Gaurav,
>
>     Looks like you are asking the routers to forward looking at the
>     second innermost label and not the topmost label. This does NOT fit
>     into the MPLS architecture. I am not sure it fits SR-IPV6
>     architecture or not, but I doubt.
>
>     Looks like your requirement is that each node on shortest path to
>     the final destination (indicated by the bottom-most Node-segment)
>     provide some service. In this regard, can you be specific about
>     wether all the nodes will provide the same service or different
>     service? It does not make sense to me for all the transit nodes to
>     execute the same service on the packet. So if they are not required
>     to provide the same service on each transit node, question is how
>     one service label will be enough to indicate which specific service
>     will need to be executed at each node.
>
>     Hope you have gone through SFC drafts already.
>
>     Thanks
>
>     -Pushpasis
>
>     *From: *spring on behalf of Gaurav agrawal
>     *Date: *Wednesday, September 9, 2015 at 6:09 PM
>     *To: *Alexander Vainshtein
>     *Cc: *"Anil Kumar S N (VRP Network BL)", "spring@ietf.org
>     <mailto:spring@ietf.org>", "mpls@ietf.org <mailto:mpls@ietf.org>",
>     Vinod Kumar S 70786
>     *Subject: *Re: [spring] [SPRING] Query related to SR Architecture
>
>     Dear Alexander,
>
>     Thanks for your inputs. Let me further elaborate on the subject.
>
>     The requirement is to make every node on the path to destination to
>     perform a specific service(Service could be anything).
>
>     Currently Service label can only follow a Node Label, because of
>     which to let every node perform same service, SR Label stack expects
>     to have node and service label for each transit node, this results
>     in huge label stack.
>
>     If we can push a service label prior to node label & each
>     intermediate node can perform below operation:
>
>     1) Pop Service Label & perform/schedule the service.
>
>     2) Decide the further forwarding based on Node Label
>
>     3) Push the service label back to stack.
>
>     With this we needn’t repeat the service label for each transit node
>     thereby making the SR Label stack COMPACT.
>
>     We can derive many optimized implementation by having this.
>
>     So, we would like to hear from You and MPLS/SPRING community about
>     our view point.
>
>     Thanks and Regards,
>
>     Gaurav Agrawal
>     Company_logo
>
>     Mobile: +91-7838700296
>     Email: gaurav.agrawal@huawei.com <mailto:gaurav.agrawal@huawei.com>
>
>     Huawei Technologies Co., Ltd.
>
>     *From:*Alexander Vainshtein [mailto:Alexander.Vainshtein@ecitele.com]
>     *Sent:* Wednesday, September 09, 2015 5:01 PM
>     *To:* Gaurav agrawal
>     *Cc:* Anil Kumar S N (VRP Network BL); Vinod Kumar S 70786;
>     spring@ietf.org <mailto:spring@ietf.org>; mpls@ietf.org
>     <mailto:mpls@ietf.org>
>     *Subject:* RE: [SPRING] Query related to SR Architecture
>
>     Gaurav,
>
>     Not sure I understand the context for your requirement.
>
>     But to the best of my understanding your requirement does not match
>     MPLS architecture.
>
>     Regards,
>
>     Sasha
>
>     Office: +972-39266302
>
>     Cell:      +972-549266302
>
>     Email: Alexander.Vainshtein@ecitele.com
>     <mailto:Alexander.Vainshtein@ecitele.com>
>
>     *From:*spring [mailto:spring-bounces@ietf.org] *On Behalf Of *Gaurav
>     agrawal
>     *Sent:* Wednesday, September 09, 2015 1:38 PM
>     *To:* spring@ietf.org <mailto:spring@ietf.org>
>     *Cc:* Anil Kumar S N (VRP Network BL); Vinod Kumar S 70786
>     *Subject:* [spring] [SPRING] Query related to SR Architecture
>
>     Hi,
>
>     We would like to have a label stack with only two labels such a way
>     that service label is a top label  and the bottom label would be SR
>     destination node label. This is to make sure each intermediate node
>     perform the specified service based on the top label while reaching
>     the destination.
>
>     We would appreciate if anyone could clarify whether SR architecture
>     could allow a service label to be a top label in a label stack.
>
>     Thanks and Regards,
>
>     Gaurav Agrawal
>     Company_logo
>
>     Mobile: +91-7838700296
>     Email: gaurav.agrawal@huawei.com <mailto:gaurav.agrawal@huawei.com>
>
>     Huawei Technologies Co., Ltd.
>
>
>
> _______________________________________________
> mpls mailing list
> mpls@ietf.org
> https://www.ietf.org/mailman/listinfo/mpls
>