Hi Acee, The ELC is only used to indicate that the advertising router as an egress of a given LSP could pop the EL/ELI pair. As such, the ingress of an LSP doesn’t know whether those transit LSRs along this LSP could perform EL-based load-balancing when inserting EL/ELI.
Best regards, Xiaohu 发件人: Acee Lindem (acee) [mailto:[email protected]] 发送时间: 2016年11月22日 1:41 收件人: [email protected] 抄送: OSPF WG List; [email protected]; Carlos Pignataro (cpignata); Xuxiaohu 主题: Re: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Bruno, From: Bruno Decraene <[email protected]<mailto:[email protected]>> Date: Monday, November 21, 2016 at 12:21 PM To: Acee Lindem <[email protected]<mailto:[email protected]>> Cc: OSPF WG List <[email protected]<mailto:[email protected]>>, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>>, "Carlos Pignataro (cpignata)" <[email protected]<mailto:[email protected]>>, Xiaohu Xu <[email protected]<mailto:[email protected]>> Subject: RE: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Acee, From: Acee Lindem (acee) [mailto:[email protected]] Sent: Monday, November 21, 2016 6:02 PM To: DECRAENE Bruno IMT/OLN; Xuxiaohu Cc: OSPF WG List; [email protected]<mailto:[email protected]>; Carlos Pignataro (cpignata) Subject: Re: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Bruno, From: Bruno Decraene <[email protected]<mailto:[email protected]>> Date: Monday, November 21, 2016 at 9:43 AM To: Acee Lindem <[email protected]<mailto:[email protected]>>, Xiaohu Xu <[email protected]<mailto:[email protected]>> Cc: OSPF WG List <[email protected]<mailto:[email protected]>>, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>>, "Carlos Pignataro (cpignata)" <[email protected]<mailto:[email protected]>> Subject: RE: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Acee, From: Acee Lindem (acee) [mailto:[email protected]] Sent: Saturday, November 19, 2016 12:33 AM To: DECRAENE Bruno IMT/OLN; Xuxiaohu Cc: OSPF WG List; [email protected]<mailto:[email protected]>; Carlos Pignataro (cpignata) Subject: Re: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Bruno, From: OSPF <[email protected]<mailto:[email protected]>> on behalf of Bruno Decraene <[email protected]<mailto:[email protected]>> Date: Friday, November 18, 2016 at 11:30 AM To: Xiaohu Xu <[email protected]<mailto:[email protected]>> Cc: OSPF WG List <[email protected]<mailto:[email protected]>>, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>>, "Carlos Pignataro (cpignata)" <[email protected]<mailto:[email protected]>> Subject: Re: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Xiaohu, Please see inline [Bruno] From: Xuxiaohu [mailto:[email protected]] Sent: Monday, November 14, 2016 1:00 PM To: DECRAENE Bruno IMT/OLN Cc: OSPF WG List; Carlos Pignataro (cpignata); [email protected]<mailto:[email protected]> Subject: ??: [OSPF] ??: WG Last Call for "Signalling ELC using OSPF" Hi Bruno, Could you please explain why the defination of the RLDC should be specific to the LB behavior of the transit LSR? [Bruno] The whole purpose of EL and ELC is to improve load balancing of MPLS packets on transit LSR. According to §6 of your draft, RLDC is also used to improve load-balancing : “The RLDC is used by ingress LSRs to determine whether it's necessary to insert an EL for a given LSP tunnel in the case where there has already been at least one EL in the label stack. > What would be the point for the ingress to add an additional EL, within the RLDC of LSR A, if LSR A do not use this EL to improve the load balancing? cf my example below where a LSR can read 5 labels, yet do not use those 5 labels for the load-balancing hence would not benefit from adding an EL within those 5 labels. BTW, it would be useful for the discussion if you could reply to the content of my email sent: Monday, November 14, 2016 (also included below). As this was already the second time I send this example on the OSPF mailing list. If I understand correctly, it seems that the text proposed by you conflict with Acee's take (see blow): " 1. The standards track IGP drafts should have a precise definition of RLD and so not require a normative reference to the MPLS entropy draft (which is informational). The IGP drafts need not precisely specify how the information is used - this can be specified via a reference to the MPLS draft. 2. The MPLS draft should precisely specify the initial use case of entropy label insertion at the ingress of the LSP. It should not limit the applicability of RLDC to this use case. " [Bruno] I’m not seeing any conflict. I agree with both points. In this thread, I’m working on 1, i.e. having a clear definition of RLD. But I would also like that this RLD advertisement be effective in improving the load-balancing of MPLS packets. I think Readable Label Depth (RLD) should be independent of EL Capability (ELC). It allows advertisement of the the maximum number of labels an OSPF router will examine in a received MPLS encapsulated packet. If an OSPF Router supports ELC, it would imply that it support the EL Capability within RLD labels. [Bruno] Would work for me, assuming that this is stated in the document, and :s/support the EL Capability within RLD labels/for load-balancing purpose, use the EL within RLD labels. I would propose the following text: “RLDC is the maximum number of labels, from the top of the stack, where the MPLS transit LSR searches for the ELI,EL pair and load-balance based on the EL if present.” I would completely decouple the two capabilities. Here is the text I would recommend. The Readable Label Depth (RLD) is the maximum number of labels, starting with top or first label in the stack, that an LSR can examine in a received MPLS packet. The supported RLD can be important when searching for an entropy label for purpose of load-balancing as the <ELI, EL> pair must be included in the first RLD labels in the stack. [Bruno] What would you use this RLD information for? Current OSPF draft proposes to use it to make a decision on whether/where adding an additional ELI, EL pairs in the stack of labels. But if we don’t know whether that additional EL, within the RLD, will be used for load-balancing purpose, the ingress can’t know whether adding this ELI, EL is useful or not. Since we are in a context where the number of labels that can be pushed is limited, we may be wasting 2 label push for zero benefit. As the name would imply, Entropy Label Capability (ELC) would be used to determine if the LSR supports load-balancing. The RLD capability, also as as the name would imply, would be used solely to determine the number of labels an LSR will read. Am I the only one who thinks this is obvious? think ELC should be defined in RFC 6790 and the SPRING Entropy label draft as opposed to the IGP advertisement drafts. [Bruno] I tend to agree that the definition of RLD, or the load-balancing behavior of a transit LSR supporting EL, would be better specified by the MPLS WG. Then the value advertised by control plane protocols/signaling. https://tools.ietf.org/html/rfc7325#section-2.4.5 talks about this, but the document is informational, and the text is a bit too large/ open to have the LSR behavior advertised in the IGP using a single integer. But this option may delay a lot the IGP draft, unless it is splitted in 2 parts (as ELC is ready). Alternatively, I’m ok with your above proposition. Why can’t we simply use the definition of entropy processing included in RFC 6790 section 4? [Bruno] Basically, section 4.3 says that the LSR can load-balance based on whatever it wants, and not necessarily based on the EL label. If it does not use the EL label, there is no point in the ingress trying to add additional ELI,EL labels within the RLD of this LSR. I’d say that a router supporting ELC recognizes the <ELI, EL> label in the first RLD labels and MAY load-balance as described in RFC 6790. An OSPF document should not update the MPLS data plane behavior. Which is the purpose of advertising the RLD (AFAIK). It is one usage of RLD �C however, the maximum number of labels an LSR can examine must not be inextricably tied to the Entropy Label processing. Thanks, Acee Thanks --Bruno Thanks, Acee _________________________________________________________________________________________________________________________ Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration, Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci. This message and its attachments may contain confidential or privileged information that may be protected by law; they should not be distributed, used or copied without authorisation. If you have received this email in error, please notify the sender and delete this message and its attachments. As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified. Thank you.
_______________________________________________ OSPF mailing list [email protected] https://www.ietf.org/mailman/listinfo/ospf
