Network Working Group Ben Abarbanel Internet Draft IPOptical, Inc., Document: draft-abarbanel-idr-bgp4-te-01.txt Senthil Venkatachalam Expiration Date: September, 2000 Alcatel, U.S.A BGP-4 support for Traffic Engineering Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026 []. This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026 except that the right to produce derivative works is not granted. This document is an Internet-Draft and is NOT offered in accordance with Section 10 of RFC2026, and the author does not provide the IETF with any rights other than to publish as an Internet-Draft Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet- Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract Currently, constraint routing (CR) and traffic engineering (TE) models do not take into consideration the big picture view of IP traffic traversing multiple autonomous systems (AS). Most of the traffic and constraint routing is based on IGP protocols such as OSPF/ISIS, etc. The resulting view of the Internet is limited to one autonomous system and areas or systems within it. Hence, the routing/forwarding functions do not select the optimum path for packets that need to traverse several autonomous systems. The proposal in this draft is that the BGP protocol can be utilized to choose the best BGP routes based on traffic engineered (TE) constraint weights. This information can be propagated between all BGP peers and calculated by the BGP AS border routers before it is deployed to their forwarding tables. draft-abarbanel-idr-bgp4-te-01.txt [page 2] 1. Conventions used in this document The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC-2119 []. 2. Overview The Internet is composed of many autonomous systems(AS). BGP acts as the road map for selecting the best paths across these systems. Today's Internet infrastructure does not consider the Internet as one homogenous system but a collection of systems. Where each system is managed autonomously by an (ISP) provider. These providers tend to filter or block routes from each other. This in itself creates a non-optimum routing. Constraint based routing adds a level of traffic influence by defining the path (all the hops along the way) that a packet should take. It does this very well within an autonomous system and within the limitation of an OSPF/ISIS areas/levels. The area border routers inside the AS will need to compute the best path to other areas until the packet arrives at its destination within the autonomous system or leaves the AS to the next Autonomous System. The choice of which AS to select is based purely on a one dimensional level, by looking at the minimum number of AS hops to a given destination or providing various weights and preferences, that are local to the AS. The BGP Multi Exit Discriminator (MED) provides some level of inter-AS metric but is still limited to the adjacent AS. The approach discussed here is to view the Internet as one homogenous entity with many AS's. Each of the AS's can have a summary weight assigned to it based on a given traffic engineering criteria. This weight can be a type of quality of service, such as maximum IGP bandwidth available, maximum number of IGP hops, maximum IGP delay across the AS from one IBGP router to the other IBGP router, maximum bandwidth for a service class A, B, C, etc. These criteria levels are summarized by the IGP for a given network destination in a given AS and exported to BGP. The summarization within an AS can be achieved using IGP constraint based routing algorithms used within link state protocols such as OSPF and ISIS. In the case of OSPF, a new opaque LSA is defined to propagate the summarized weights between OSPF areas. This draft is organized as follows. Section 3 details the approach at the BGP level between ASs and Section 4 discusses the approach at the IGP level within an AS. The various traffic engineering weights are discussed in Section 5. The following sections deal with the redistribution of the traffic engineering weight information from IGP to BGP, configuration issues, route flapping, and ISP issues. The BGP and IGP approaches together will provide a comprehensive traffic engineering routing system across the internet. draft-abarbanel-idr-bgp4-te-01.txt [page 3] 3. BGP Level The method used to propagate constraint summarization weights for each AS is to define a new attribute which contains QOS like sub fields that can include such parameters as bandwidth, number of hops, delay, various QOS service classes, and so on. Historical NLRI information is contained as well. These sub fields will be termed TE weights and each BGP router would propagate this information to its peers. Making it possible for any router that handles the data to have a traffic engineering view of all the paths and their associated TE Weights for a given route to a given router. When the BGP RIB database is loaded with TE Weight information, a TE capable BGP router would compute based on TE manual configurations criteria the best BGP route for a given destination. The BGP Route Selection process is extended to support a TE way of prioritizing the best routes for any configured destination. In which the order and preference of the routes can be changed to give the TE weight attribute a higher priority than other attributes. Different TE Weight types could be manually assigned a different level of priority in order to strategize a system where the BGP route selection criteria could be further optimized. 3.1 BGP Routes Originating in Local AS Routes that originated within the same AS will not be calculated for TE weights since there is no BGP optimization that can be achieved. In this case, all TE Weight calculations are strictly done at the IGP level. As a rule, IBGP routers will not propagate to other IBGP routers the new TE Weight attribute for routes that originated in the same AS. Routes that came from other EBGP peers could contain the new TE Weight attribute which will be propagated to IBGP peers. 3.2 Route Reflector Functionality When a Route Reflector receives an update messages with the Aggregated TE Weight attribute it will simply add it to the BGP RIB-IN/OUT for local use and propagate it to other IBGP peers without modifying any of its fields. 3.3 When to Add TE Weight to BGP Update messages When an IBGP router receives an update message from another IBGP router for a (route) destination outside its own AS, it will consult the local FIB database for its reachability via the corresponding update message BGP Next Hop. Using this BGP Next Hop it will check the associated TE summary weight provided in the FIB by the extended IGP constraint routing calculations. It will add or compare the IGP TE summary weight with the Aggregated TE Weight that was propagated in the update message and create the next level Aggregated TE weight. Next, it will add the Aggregated TE weight to the local BGP RIB and use it for the BGP route selection process. In addition, the current router will propagate the new Aggregated TE weight value to all of its EBGP peers except the one that introduced the route. draft-abarbanel-idr-bgp4-te-01.txt [page 4] Propagation, aggregating BGP TE Weights, and deploying into the BGP RIB and FIB tables is performed from one router to the next such that the only TE weight that is detected by any router along an AS path is accurate enough to cover the TE measurements from a given router to the route's point of origin. 3.4 Route Aggregation Impacts As routes are propagated from one AS to the next, it is possible that providers have configured their routers to aggregate more specific routes into general ones. This condition will cause the specific routes and their TE Weight attributes to get lost or reduced in accuracy. In order to keep track of this information, it is proposed that the specific routes and their TE Weights be contained in the new TE Weight attribute fields. It would be possible at any router along the route propagation paths to determine the detail accuracy of the TE weights that made up the super aggregate TE Weight. Example 1: A single AS-Path with no route aggregation, and TE weights are aggregated. In AS1 In AS2 Arriving in Dest AS3 N1N2,AS1,TE1 ----> N1N2,AS2,(TE1+TE2) ----> N1N2,TE12 In AS3 all we need is the routes(N1N2) and the TE weight TE12, which is the aggregate of TE1 and TE2. Example 2: Aggregating routes from multiple AS-Paths, and TE weights are super aggregated too. In AS1 In AS2 In Dest AS3 N1N2,AS1,TE1 ----> N1N2,AS2,(TE1+TE2) --> N1N2,TE12 --> see below In AS4 In AS5 In Dest AS3 N3N4,AS5,TE5 ----> N3N4,AS5,(TE5+TE6) --> N3N4,TE56 --> see below At this point in AS3, routes N1N2 and N3N4 are aggregated together to form super aggregate route N13N24, with the aggregated TE weight of (TE12+TE56) = Super aggregated TE WEIGHT TE1256. In order to keep track of which routes made up TE12 and TE56 we keep two TE Weight paths in the super aggregated route N13N24 as follows: N13N24/TE1256 = (N1N2/TE12, N3N4/TE56) in the TE Weight Attribute field. This list of historical routes can be as large as the number of route aggregation points that are performed along the AS-Paths. *Note: For each list of historical routes and TE weights, there will be one or several TE Weight types. See section 3.5 for details. draft-abarbanel-idr-bgp4-te-01.txt [page 5] 3.5 TE Weight Calculation Impacts As was seen in section 3.4, routes and TE Weights are aggregated together into less specific information and as a result the TE weights become more and more global. This affect creates sub optimal data for choosing the best path across a series of ASs. If there is an historical trace of route and TE weights that created the global aggregated route, than its possible to make more accurate decision in choosing the best BGP route. This is done during the route selection process which is run for the same route from multiple BGP peers. With the information in the TE Weight Attribute its possible to look inside an aggregated route and more precisely see its composition (sub-aggregation of routes and more specific TE weights) and determine the best TE route and related BGP Next Hop. 3.6 TE Weight Attribute Format The BGP update message will contain a new Optional Transitive attribute called TE Weight (type code ?). This contains one or several aggregated TE Weight types and their historical sub aggregated routes and TE weights. The amount of aggregation would depend on the number of AS's the route had traversed from the point of origin to the current router. A TE Weight type must have the same consistency or granularity throughout all the routers that have the software to aggregate, calculate, and propagate it further. |-- Optional(1) or well known(0) | |-- transitive(1) or non-transitive(0) | | |-Attr Flags | | | 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-|-|-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |1|1| FLAGS | Attr Type (?) | Number TE Weight Lists | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ +==============|===============|===============|===============+ + 1st Super Aggregated TE Weight list entry | +==============|===============|===============|===============+ | Number of TE | TE Weight | 1st Super Aggregated TE Weight| | Weight types | Type 1 | Value | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ | 2nd TE Weight | 2nd Super Aggregated TE Weight |nth TE Weight| | Type | Value |type | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |nth Super Aggregated TE Weight| Number of Aggregated Route | | Value | Prefixes | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |route prefix 1| 1st Aggregated IP Route prefix | |length | 1st, 2nd, 3rd byte | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |1st route pref| route prefix n| nth Aggregated IP route prefix| | 4th byte | length | 1st, and 2nd byte | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |nth Aggregated IP route prefix| | | 3rd and 4th byte | ... | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ draft-ietf-abarbanel-bgp4-te.txt [page 6] 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ +==============|===============|===============|===============+ + nth Super Aggregated TE Weight list entry | +==============|===============|===============|===============+ | Number of TE | TE Weight | 1st Super Aggregated TE Weight| | Weight types | Type 1 | Value | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ | 2nd TE Weight| 2nd Super Aggregated TE Weight |nth TE Weight | | Type | Value |type | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |nth Super Aggregated TE Weight| Number of Aggregated Route | | Value | Prefixes | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |route prefix 1| 1st Aggregated IP Route prefix | |length | 1st, 2nd, 3rd byte | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |1st route pref| route prefix n| nth Aggregated IP route prefix| | 4th byte | length | 1st, and 2nd byte | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ |nth Aggregated IP route prefix| | | 3rd and 4th byte | ... | +-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-++ - Each TE Weight type could be: o Maximum Bandwidth Available o Maximum Number of IGP Hops o Maximum Transit Delay o Color o Etc. - Each TE Weight value can be in units relevant to its use: o For maximum bandwidth could be Megabits/second. o For Maximum transit delay could be in milliseconds. - Aggregated IP Route Prefix: Defines the aggregation of several routes into a single route for a given AS. - Route Prefix Length: The number of significant bits from the left side or high order byte of the IP address. As defined in CIDR. Example: IP address 10.20.30.40, High order Byte is the 10. - Super Aggregated TE Weight: Each TE weight will have a different algorithm for aggregation and meaning. See section 5.0 for details. 3.7 Phasing TE Weights The use of TE weights can be manually configured on a destination route prefix basis, and thus its possible to enable/disable TE Weights on any BGP route. The functionality can be easily phased into an AS where some BGP routers are TE Weight capable and others are not. The new attribute will be Optional and Transitive which implies that older BGP routers will be required to propagate the new attribute in update messages to their peers. draft-abarbanel-idr-bgp4-te-01.txt [page 7] When not all of the BGP routers along a given TE Path support the TE Weight attribute, the TE weight calculations is not done there and the best BGP route selection computation will be less than optimal. 4. IGP TE Constraints based Calculations and AS Summarization for TE The IGP level routing protocol used is generally either OSPF or ISIS. The basic OSPF and ISIS protocols allow routing based on a static metric. These protocols can be extended to take into consideration various traffic engineering criteria such as unreserved bandwidth, delay, colors, OSPF metric, etc. The approach detailed below is oriented more toward OSPF, but will be readily extended to ISIS. If the IGP inside a routing domain is OSPF, the AS is divided into areas - each of which is a collection of routers and networks. The routers within an area know the complete topology of the area by means of LSA database synchronization with their neighbors at all times. Through the use of the opaque LSAs [8], it is possible to maintain the traffic engineering topology (in addition to the regular network topology) and hence perform route calculations based on traffic engineering criteria within an area inside an AS that runs OSPF. However, this restricts the traffic engineering calculations to within an area. To determine the traffic engineering weight to any network or router in the AS, the approach proposed for OSPF is to define a new opaque LSA for traffic engineering that summarizes the traffic engineering weights of every network from the perspective of an area border router (ABR), for each area in the AS. This traffic engineering summary LSA is analogous to the summary LSA in OSPF. More details on this work can be found in [9]. The destination network could either lie in the same area as the ASBR, or in a different area. If the destination network and the ASBR are in the same area, the opaque traffic engineering LSAs as defined in [8] for the area will suffice to calculate the traffic engineering metric to the destination network. If the destination network lies in a different area than the ASBR, the weight is a combination (such as simple addition, or max) of the weight to the ABR and the weight described in the traffic engineering summary LSA originated by the ABR that contains the destination network. In this manner, the traffic engineering weights for all the networks in an AS can be computed. Hence the ASBRs will also be able to determine the aggregate traffic engineering weights across the AS to other ASBRs and use this in the BGP advertisements. 5. Weights The traffic engineering weights act as a cost or distance function, describing the quality of a path to a destination network in traffic engineering terms. The traffic engineering weights currently proposed are: draft-abarbanel-idr-bgp4-te-01.txt [page 8] 1. Available Bandwidth (in bits/sec) 2. Unreserved Bandwidth (in each of several classes) 3. Colors (or class types) (8 types) 4. Transit Delay (in milliseconds) 5. IGP Metric 6. IGP Hops These traffic engineering weights can be dynamically measured or statically configured for each interface in a node. The weights can be propagated within an IGP area using the opaque traffic engineering LSA. The area border routers (ABRs) can summarize the traffic engineering weights to destination networks in their area and flood this information into the backbone through the use of the traffic engineering summary LSA as proposed in [9]. The summary LSAs can then be flooded by other ABRs into their areas. Any ASBR can determine the TE weight to a destination network by examining its TE database and performing a simple dijkstra like calculation by its associated IGP. This calculation can be a real dijkstra if the weight is additive (delay, or hops or IGP metric), or dijkstra like if the weight is a minimum to maximum constraint (available bandwidth, unreserved bandwidth, and colors). Depending on the level of processing desired, the number of supported TE weights should be configured. These calculations result in a value for each TE Weight for each destination network in the AS and for each ASBR across the AS. The weights from the IGP are then exported into BGP for propagation to its EBGP peers. The following two subsections deal specifically with the treatment of the available bandwidth and delay TE weights, during the calculations: Maximum Available Bandwidth: IBGP/IGP router within a given AS will pick the smallest available Bandwidth link from the ingress to the egress of the IBGP to IBGP path and use that value as the summary TE weight. Next the egress IGP will export this information to BGP which will use and propagate it to its EBGP peers as the first aggregated TE Weight value. As the route is propagate to each IBGP peer in another AS, the Aggregated TE weight value in the message is compared with the summary TE weight for the BGP Next Hop router defined in the message. Whichever value is the smaller of the two, will be considered the next level of aggregation value and propagated further to the next AS. This mechanism will continue till the route reaches the furthest points in the Internet. This way the smallest available bandwidth value will be used as the overall Aggregated TE Weight value for any given route along the TE path. Maximum Delay: IBGP/IGP router within a given AS will obtain the delay from the ingress to the egress of the IBGP to IBGP path and use that value as the summary TE "delay" weight. Next the egress IGP will export this information to BGP draft-abarbanel-idr-bgp4-te-01.txt [page 9] which will use and propagate it to its EBGP peers as the first aggregated TE Weight "delay" value. As the route is propagate to each IBGP peer in another AS, the Aggregated TE weight value in the message is added with the summary TE "delay" weight for the BGP Next Hop router defined in the message. The sum of the two is considered the next level of aggregation and propagated further to the next AS. This mechanism will continue till the route reaches the furthest points in the Internet. *Note: See section 3.4 for Route Aggregation Impacts which must be followed in order for this mechanism to be optimal. As mentioned in section 3.4, historical routes/TE Weights are carried in this new attribute for routes that are aggregated several times. 6. TE Weights Redistributed from the IGP to BGP In order for a given router to inherit the TE weights from IGP, a mechanism must be provided to do that. What is proposed is that the standard FDB database contain TE summary weight fields based in each route entry, with their TE weight type, TE weight value and associated AS number. In other words, in all the BGP routers, the running IGP will inject the TE summary weight information into the FDB database and signal BGP to import it to its BGP RIB database for propagation to other BGP peers. 7. Configuring the TE feature for BGP and IGP. In order to minimize the amount of information produced by the TE weight parameters across the entire Internet as described in section 3.6, it is recommended that the TE weight be configured on at the network route's point of origin. Implying that the AS router that originated the route into BGP will add the TE Weight attribute and all transit routers that propagate this route will perform the TE Weight processing when they see the TE Weight Attribute. This way it will be possible to limit the amount of TE Weight information being propagated across the Internet. What is needed to manually configure the BGP Best Route Selection criteria: I. To Enable TE Weight Feature for a given network prefix. a. Define each network prefix for TE Weight functionality. Once enabled all transit BGP TE routers will process the TE Weight aggregation and propagate the TE Weight attribute to their peers. II. To manually configure the BGP Best Route Selection Criteria in each BGP TE Router, configure the following parameters: a. TE Weight Type o Maximum Bandwidth Available (Megabits/sec) o Unreserved Bandwidth (in each of several classes) o Maximum Number of IGP Hops o IGP Metric o Maximum Transit Delay (in Milliseconds) o Color (or class types) (8 types) draft-abarbanel-idr-bgp4-te-01.txt [page 10] b. TE Weight Priority 1st choice = Make this TE Weight the first choice for best BGP route selection criteria 2nd choice = Make this TE Weight the 2nd choice for best BGP route selection criteria nth choice = Make this TE Weight the 2nd choice for best BGP route selection criteria c. Example of BGP TE Topology AS2 +---------+ 70 MB +--------------+ | R6+------------------------------------| R7--net1--R13| | AS5 ||70MB |/ \ 80MB| | | R5| / \ | | |---------+ /| \30MB | | \ 30MB / | \------R8| \70MB / +------------+-+ \ AS1 / | +-------+--------|----------/+ |30MB | R2 | 30MB | R3|\ +------+-------+ | | /-|--------|---------/ | \ | R9-net2 | | | / | | | \60MB | 60MB / | Source--+->R1/ | Area 0 | Area 2 | \ | / | | \----+--------+--------\ | \ | / | | Area 1| 30MB | R4 | \---+-R10 AS3 | +-------+--------|-----------+ +-- -\---------+ \ \ \30MB \30MB \ \ \+----------- --\----+ |R11----------- R12 | | 30MB / | | / | | AS4 net3 | +-------------------+ Figure 1. BGP TE Topology draft-abarbanel-idr-bgp4-te-01.txt [page 11] Table 1. Router R1 BGP TE Weight Table +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+++-+-+-+-+-+-+-+-+-+-+-+-++ |Entry # | Route | BGP Next Hop | Aggregated TE Weight | | | | | Bandwidth | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+++-+-+-+-+-+-+-+-+-+-+-+-++ | 1 | NET1 | R7 via R3 | 30 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 2 | *NET1 | R5 via R2 | 70 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 3 | NET1 | R10 via R3 | 60 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 4 | NET1 | R11 via R4 | 30 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 5 | NET2 | R11 via R4 | 30 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 6 | *NET2 | R10 via R3 | 60 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 7 | NET2 | R7 via R3 | 30 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- | 8 | NET2 | R5 via R2 | 30 MB | +-+-+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+|+-+-+-+-+-+-+-+-+-+-+-+-+- *Note: Although we show only one TE weight type under each AS subfield it is possible to have multiple TE weight types and have them prioritized by manual configurations in each BGP TE router. In Figure 1 we see that R1 is the current router receiving Update messages from R2, R3, and R4. All the Update message contain the new BGP TE Weight Attribute with the weights as shown in Table 1. From this data, we see that for NET1, Entry #2 is the best choice amongst other NET1 entries, because it guarantees that we get 70MB bandwidth from AS2 and AS5. Assumption, operator had configured the TE (bandwidth) weight as the highest priority criteria. Also for NET2 entries, entry #6 is the best choice, even though it includes more AS hops, but it guarantees that at least 60MB of bandwidth will be available. This case shows, how sometimes sacrificing more AS hops for guaranteed bandwidth service could work. 8. Route Flapping and Impacts on BGP Aggregated TE Weights In the real world route flapping is a normal occurrence. It is expected that the router performing TE Weight Aggregation will be notified by the IGP that routes have gone down/up and it will be required to recalculate the aggregated TE Weight. Obviously, when a new aggregated TE weight is defined for an aggregated route, the router performing this calculation will be required to deploy it to its BGP peers. draft-abarbanel-idr-bgp4-te-01.txt [page 12] 9. BGP Confederations Since the IGP domain will cover the entire AS, any mapping of BGP Confederation within each of the sub-ASs will not be used for BGP TE aggregation. Internal Sub-AS BGP routers running either EBGP or IBGP will not aggregate the new TE Weight attribute since that job will only be done By the BGP ASBR routers. They will however, use these attributes in their Route selection criteria and propagate them further to all inter/intra sub- AS peers. 10. ISP Issues and Dependencies TBD 11. Conclusion As has been shown, the Internet is a complex WEB of systems within systems. Traffic Engineering is the only solution that attempts to control/contain the explosion of resources within routers. The technique described in this manuscript defines a way for BGP routers to influence and control the routing paths chosen for routes that are bound by Traffic Engineering Weights. Thereby, enabling the Inter-AS providor a more uniform method of allocating resources and providing gauranteed service using BGP TE across multiple Autonomous Systems. 12. Security Considerations The new BGP TE Weight attribute does not change the underlying security issues inherent in the existing BGP-4 design. 13. References [1] "A Border Gateway Protocol 4 (BGP-4)", Y. Rekhter & T. Li, RFC1771, March 1995 [2] Awduche, D., Malcolm, J., Agogbua, J., O'Dell, M., McManus, J., "Requirements for Traffic Engineering Over MPLS", RFC 2702, September 1999. [3] Awduche, D., Rekhter, Y., Drake, J., Coltun, R., "Multi-Protocol Lambda Switching: Combining MPLS Traffic Engineering Control With Optical Crossconnects", draft-awduche-mpls-te-optical-01.txt. [4] Jamoussi, B. "Constraint-Based LSP Setup using LDP", Work in Progress, Internet Draft , September 1999. [5] Braden, R., Zhang, L., Berson, S., Herzog, S., "Resource ReSerVation Protocol (RSVP) -- Version 1 Functional Specification", RFC 2205, September 1997. [6] Awduche, D. et al "Extensions to RSVP for LSP Tunnels", Work in Progress, Internet Draft September 2000. [8] Katz, D. and Yeung D., "Traffic Engineering Extensions to OSPF", Internet Draft [9] Venkatachalam, Senthil and Abarbanel, B., "Traffic Engineering Summary Extensions to OSPF", work in progress. [10] Villamizar, "BGP Route Flap Damping", RFC2439, November 1998 14.Acknowledgments To be supplied in future revisions. 15. Author's Addresses Ben Abarbanel IPOptical, Inc. 11480 Sunset Hills Rd, Suite 200E Reston, VA 20190 email: ben.abarbanel@ipoptical.com Senthil Venkatachalam Alcatel USA 45195 Business Court, Suite 400 Dulles, VA 20166 email: senthil.venkatachalam@usa.alcatel.com