Hello, I have been selected as the Routing Directorate reviewer for this draft. The Routing Directorate seeks to review all routing or routing-related drafts as they pass through IETF last call and IESG review, and sometimes on special request. The purpose of the review is to provide assistance to the Routing ADs. For more information about the Routing Directorate, please see ​http://trac.tools.ietf.org/area/rtg/trac/wiki/RtgDir Although these comments are primarily for the use of the Routing ADs, it would be helpful if you could consider them along with any other IETF Last Call comments that you receive, and strive to resolve them through discussion or by updating the draft. Document: draft-ietf-bess-dci-evpn-overlay-08 Reviewer: Alexander (“Sasha”) Vainshtein Review Date: 14-Feb-18 IETF LC End Date: 09-Feb-18 Intended Status: Standards Track Summary: I have some minor concerns about this document that I think should be resolved before publication. Comments: The document is well written, but understanding of both the EVPN technology (RFC 7432) and network virtualization basics are mandatory prerequisites for the readers. My personal expertise in both these areas is limited, and this may affect the quality of the review. From my POV this draft complements the EVPN Overlay draft (already approved by the IESG for publication as an RFC): it discusses interaction between the EVPN Overlay within the DC with some options that implement L2 connectivity in the WAN that provides the infrastructure for the DC interconnect. I have identified two groups of specifications in the draft: - Specifications in the first group explain how the mechanisms already defined in other specifications (mainly in RFCF 7432) should be used to provide DCI Interconnect that uses EVPN as the overlay within the DC. One example can be found in Section 3.5.2 that recommends usage of ARP/ND Proxy cache in the DC Gateways to prevent flooding of ARP/ND messages within the DC, many other examples can be added - Specifications in the second group define new behavior. One example is the proposed (in Section 3.5.1) usage of the Unknown MAC Route (UMR) to prevent overwhelming the NVEs with the need to learn zillions of MAC addresses in the remote DCs. As part of preparation of this review I have discussed my comments with the authors who have been most responsive and cooperative - so much so that they have addressed some of my earlier comments in the latest (-08) version of the draft. As a consequence, I had to remove the already addressed comments from the final version of my review, and to ask the authors not to post a new version before posting of the review. I would like to express my gratitude to the authors and, especially, to Jorge for excellent cooperation. Major Issues: None found. Minor Issues: 1. From my POV this draft should be marked as updating RFC 7432 in its metadata. The update should refer to several aspects including at least the following: a. Use of Ethernet PWs (see Figure 1 in the draft) as an Ethernet Segment. RFC 7432 defines Ethernet Segment as a set of Ethernet links that connect a customer sit to one or more PEs. b. Use of the Unknown MAC Route (UMR). RFC 7432 only says that a PE may flood unicast frames with unknown destination MAC to all other PEs but does not have to do that, with the decision being a matter of local policy; it neither defines any mechanisms that advertise a specific PE and a specific Ethernet Segment attached to this PE as the “default next Hop” for all unknown destination MAC addresses, nor prevents usage of such mechanisms. 2. Definitions of VLAN-based and VLAN bundle-based Ethernet Segments in RFC 7432 do not cover the case of PW hand-off between the WAN and DC GW in the Decoupled model. While this looks like a straightforward extension, it should be clarified in the draft IMO. 3. The UMR and its encoding (defined in Section 3.5 of this draft) already have been defined in RFC 7543. I suggest to remove the UMR definition from the text and to replace it with a Normative reference to RFC 7543. At the same time RFC 7543 and this draft seem to use the UMR differently, and these differences should also be clarified in the draft. 4. The draft presents two DC Interconnect models (shown in Figure 1 and Figure 2 respectively): Decoupled Interconnect and Integrated Interconnect. These diagrams create an impression that the same model must be used in all the interconnected DCs – but this impression is wrong. Actually (clarified that with the authors) the model is a local issue between a specific DC GW and WAN, so that the same interconnect can use the Decoupled model in the GWs of some DCs and Integrated model in the GWs of other DCs. 5. The EVPN Overlay draft defines two modes for implementing DC Interconnect: using DC GWs and using ASBRs. Both models (Decoupled and Integrated Interconnect) discussed in this draft are actually sub-models of the model of DC Interconnect that uses GWs. The draft actually mentions that, but quite late - in the Security Considerations section where I, for one, would not be looking for this kind of information at all. I would suggest moving this clarification to the Introduction section and only keeping the text that deals with the security benefits of the GW-based model in Section 5. (Aside: The draft has successfully passed the SecDir review, so I hope that such a change would not cause any issues.) 6. In Section 4.2 the draft discusses Integrated DC interconnect that uses VPLS in WAN. It refers to RFC 4761, RFC 4762 and RFC 6074 for the definition of VPLS and says (in section 4.2.1) that “different route-targets for the DC and for the WAN are usually required.” Since BGP in general and RTs specifically are not relevant for VPLS based on RFC 4762, the corresponding exception should be added to the text. 7. In section 4.2.1 the draft also says that “the GWs will be provisioned with I-ESI” where I-ESI stands for the Interconnect Ethernet Segment Identifier. But this is all about the Integrated Interconnect – so what represents the Interconnect Ethernet Segment (to be identified by I-ESI) in this model? 8. Both Section 4.2.1 and Section 4.4.6 mention a local Attachment Circuit (AC) on the DC GW (in the latter case – as one of the advantages of the DC Interconnect solution that uses EVPN-MPLS in the WAN). But such ACs are not shown in the diagram depicting the Integrated model. Some clarifying text would be helpful. In particular, it would be nice to explain why local ACs are considered as benefits of the Integrated DC Interconnect solution that uses EVPN-MPLS in the WAN if they are also possible in the Integrated DC Interconnect solution that uses VPLS (at least as implied by them being mentioned in section 4.2.1). 9. Section 4.6 discussed the Integrated DC Interconnect solution that uses EVPN-VXLAN in the WAN. Other encapsulations (e.g., MPLS in GRE) are not discussed. It would be nice if the authors could clarify the reasons for including one encapsulation and excluding all others. Nits: I did not check the draft for typos. I would only like to mention that the draft mentions (in several places) “existing Technical Specifications” as if they were Royalty - looks a bit exaggerated to me. Regards, Sasha Office: +972-39266302 Cell: +972-549266302 Email: Alexander.Vainshtein@ecitele.com ___________________________________________________________________________ This e-mail message is intended for the recipient only and contains information which is CONFIDENTIAL and which may be proprietary to ECI Telecom. If you have received this transmission in error, please inform us by e-mail, phone or fax, and then delete the original and all copies thereof. ___________________________________________________________________________