90 likes | 270 Views
BGP Traffic Engineering Attribute draft-fedyk-bgp-te-attribute-03.txt. Yakov Rekhter, Don Fedyk, Hamid Ould-Brahim IETF 70 th , Vancouver Meeting, CCAMP, December 2007. Some History of BGP-TE attribute.
E N D
BGP Traffic Engineering Attributedraft-fedyk-bgp-te-attribute-03.txt Yakov Rekhter, Don Fedyk, Hamid Ould-Brahim IETF 70th, Vancouver Meeting, CCAMP, December 2007
Some History of BGP-TE attribute • IETF 67th November 2006 presented to L1VPN WG; chairs requested the draft to be presented to IDR WG. • IETF 68th (March 2007) presented to IDR WG, incorporated comments, no objections from IDR WG. • IETF 68th CCAMP chair proposed that the best home for this work is CCAMP (see IDR IETF68th meeting minutes). • And volunteered to bring the ID to CCAMP WG. • March 2007 CCAMP chair proposed to merge the draft with draft-vasseur-ccamp-ce-ce-te-03.txt (see next slide for more). • The latter draft was presented neither to L3VPN nor IDR WGs so far • BGP-TE attribute is a normative reference in L1VPN BGP auto discoverydraft. • draft-ietf-l1vpn-bgp-auto-discovery has passed L1VPN WG last call (October 2007) • September 2007 CCAMP chair suggested to re-evaluate suitability of the draft for CCAMP WG. • October 1st 2007 authors of BGP-TE attribute asked CCAMP chairs to accept the draft as a CCAMP WG document. • December 2007: CCAMP chair positions the whole problem to be just “Distribution of TE information from outside ASes”. IETF70th, Vancouver Dec 2007
Why merge is not appropriate • Merging draft-fedyk-bgp-te-attribute and draft-vasseur-ccamp-ce-ce-te is not appropriate: • The former specifies encoding of a BGP attribute. • The latter specifies one possible use of such attribute. • Without describing the format of the attribute (for future revisions of the document). • BGP attributes are defined in their own documents with pointers to documents that specify usage of such attributes. • E.g., BGP Extended Community attribute. • No compelling reasons to change this. • With this in mind we should consider discussion on merging the two drafts closed with the resolution not to merge them. IETF70th, Vancouver Dec 2007
What TE information to carry ? (1) • Debate on the CCAMP mailing list: • Agreement on the fact that TE information needs to be captured in a new BGP TE attribute. • Disagreement is mainly on whether • (a) to carry all and exactly the same TLVs as in OSPF-TE or • (b) only a (proper) subset of the information carried by all these TLVs. IETF70th, Vancouver Dec 2007
What TE information to carry ? (2) • Currently there is only one application that is an IETF WG (L1VPN WG) document that requires BGP to distribute traffic engineering information. • draft-vasseur-ccamp-ce-ce-te is a proposal with applicability to L3VPNs. • Traffic engineering information required by L1VPN forms a (proper) subset of the information carried in all the TE TLVs carried in OSPF Opaque LSA. • Moreover, some of the information carried in the TE TLVs of OSPF Opaque LSAs is carried in BGP as part of L1VPN NLRI • Carrying the same information in the BGP-TE attribute is useless • The information carried in BGP-TE attribute reflects L1VPN usage of this attribute. • Authors of draft-vasseur-ccamp-ce-ce-te want to carry all and exactly the same TLVs as in OSPF-TE. • If the intent is to carry all the OSPF-TE TLVs, as carried in OSPF opaque LSA “as is”, then why not use OSPF ? • OSPF already provides mechanism to carry opaque LSA. • Why BGP would be a better transport for opaque LSA then OSPF ? IETF70th, Vancouver Dec 2007
TE Information (OSPF for example) MPLS RFC 3630 GMPLS Depreciates two fields 1 Link type (1 octet) 2 Link ID (4 octets) 3 Local interface IP address (4 octets) 4 Remote interface IP address (4 octets) 5 Traffic engineering metric (4 octets) 6 Maximum bandwidth (4 octets) 7 Maximum reservable bandwidth (4 octets) 8 Unreserved bandwidth (32 octets) 9 Administrative group (4 octets) GMPLS RFC 4203 OR 11 Link Local/Remote Identifiers (8 octets) 14 Link Protection Type (4 octets) 15 Interface Switching Capability Descriptor (variable) 16 Shared Risk Link Group (variable) draft-fedyk-bgp-te-attribute-03.txtNew BGP attribute (BGP-TE Attribute) Interface Switching Capability Descriptor (see RFC4203, RFC4205) BGP already includes the address in NRLI (3, 4 or 11) for L1VPN BGP MED may carry the TE Metric As indicated in the mailing list other information can be added if there is WG consensus IETF70th, Vancouver Dec 2007
BGP-TE Attribute usage • The information carried in BGP-TE attribute reflects L1VPN usage of this attribute. • The authors are willing to include certain guidelines for future use of the attribute. • Provided these guidelines are developed using IETF consensus process. IETF70th, Vancouver Dec 2007
More on BGP-TE usage by L1VPN • Viewing L1VPN usage of BGP-TE attribute as just “distribution of TE information outside ASes” does not adequately reflect the application. • For one thing, L1VPN deals with multiple routing/addressing realms. • “Area” or “AS” as scopes for “flooding” TE information, while relevant in the context of distribution of TE information outside AS, is not applicable in the context of L1VPN. • PE with a given L1VPN attached to it redistributes the reachability/membership information (and the TE information as well) to all other PEs that have that L1VPN. • Has nothing to do with areas and/or ASes. IETF70th, Vancouver Dec 2007
Summary • The “Distribution of TE information from outside ASes” issues brought in CCAMP are unrelated/non-applicable to the use of BGP-TE attribute in L1VPN WG. • BGP-TE attribute is a normative reference in L1VPN BGP auto discovery draft. • draft-ietf-l1vpn-bgp-auto has passed L1VPN WG last call (October 2007). • No progress of draft-fedyk-bgp-te-attribute in CCAMP. • Lack of progress in BGP-TE in CCAMP blocks progress of draft-ietf-l1vpn-bgp-auto-discovery. • Proposal: Move draft-fedyk-bgp-te-attribute to L1VPN WG. IETF70th, Vancouver Dec 2007