Internet Draft Network Working Group Bruce Davie Internet Draft Jeremy Lawrence Expiration Date: May 1998 Keith McCloghrie Yakov Rekhter Eric Rosen George Swallow Cisco Systems, Inc. Paul Doolan Ennovate Networks, Inc. November 1997 Use of Label Switching With ATM draft-davie-mpls-atm-00.txt Status of this Memo This document is 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." To learn the current status of any Internet-Draft, please check the "1id-abstracts.txt" listing contained in the Internet-Drafts Shadow Directories on ftp.is.co.za (Africa), nic.nordu.net (Europe), munnari.oz.au (Pacific Rim), ds.internic.net (US East Coast), or ftp.isi.edu (US West Coast). Abstract Davie, et al. [Page 1] Internet Draft draft-davie-mpls-atm-00.txt October 1997 A Multi Protocol Label Switching architecture is described in [1]. Label Switching enables the use of ATM Switches as Label Switching Routers. The ATM Switches run network layer routing algorithms (such as OSPF, IS-IS, etc.), and their data forwarding is based on the results of these routing algorithms. No ATM-specific routing or addressing is needed. This document describes how the label switching architecture is applied to ATM switches. Contents 1 Introduction ........................................... 2 2 Definitions ............................................ 3 3 Special Characteristics of ATM Switches ................ 3 4 Label Switching Control Component for ATM .............. 4 5 Hybrid Switches (Ships in the Night) ................... 5 6 Use of VPI/VCIs ....................................... 5 7 Label Allocation and Maintenance Procedures ............ 6 7.1 Edge LSR Behavior ...................................... 6 7.2 Conventional ATM Switches (non-VC-merge) ............... 7 7.3 VC-merge-capable ATM Switches .......................... 9 7.4 Efficient use of label space ........................... 11 8 Encapsulation .......................................... 11 9 Security Considerations ................................ 12 10 Intellectual Property Considerations ................... 12 11 References ............................................. 12 12 Acknowledgments ........................................ 12 13 Authors' Addresses ..................................... 12 1. Introduction A label switching architecture is described in [1]. It is possible to use ATM switches as label switching routers. Such ATM switches run network layer routing algorithms (such as OSPF, IS-IS, etc.), and their forwarding is based on the results of these routing algorithms. No ATM-specific routing or addressing is needed. When an ATM switch is used for label switching, the label on which forwarding decisions are based is carried in the VCI and/or VPI fields. (It is possible to carry multiple labels in the VCI and/or VPI fields, but the scope of this document is restricted to the case Davie, et al. [Page 2] Internet Draft draft-davie-mpls-atm-00.txt October 1997 of a single label.) The characteristics of ATM switches require some specialized procedures and conventions to support label switching. This document describes those aspects of label switching which are specific to ATM. 2. Definitions A Label Switching Router (LSR) is a device which implements the label switching control and forwarding components described in [1]. A label switching controlled ATM (LC-ATM) interface is an ATM interface controlled by the label switching control component. Packets traversing such an interface carry labels in the VCI and/or VPI field. An ATM-LSR is a LSR with a number of LC-ATM interfaces which forwards cells between these interfaces using labels carried in the VCI and/or VPI field. A frame-based LSR is a LSR which forwards complete frames between its interfaces. Note that such a LSR may have zero, one or more LC-ATM interfaces. An ATM-LSR cloud is a set of ATM-LSRs which are mutually interconnected by LC-ATM interfaces. The Edge Set of an ATM-LSR cloud is the set of frame-based LSRs which are connected to the cloud by LC-ATM interfaces. VC-merge is the process by which a switch receives cells on several incoming VCIs and transmits them on a single outgoing VCI without causing the cells of different AAL5 PDUs to become interleaved. 3. Special Characteristics of ATM Switches While the label switching architecture permits considerable flexibility in LSR implementation, an ATM-LSR is constrained by the capabilities of the (possibly pre-existing) hardware and the restrictions on such matters as cell format imposed by ATM standards. Because of these constraints, some special procedures are required for ATM-LSRs. Some of the key features of ATM switches that affects their behavior as LSRs are: Davie, et al. [Page 3] Internet Draft draft-davie-mpls-atm-00.txt October 1997 - the label swapping function is performed on fields (the VCI and/or VPI) in the cell header; this dictates the size and placement of the label(s) in a packet. - multipoint-to-point and multipoint-to-multipoint VCs are generally not supported. This means that most switches cannot support `VC-merge' as defined above. - there is generally no capability to perform a `TTL-decrement' function as is performed on IP headers in routers. This document describes ways of applying label switching to ATM switches which work within these constraints. 4. Label Switching Control Component for ATM To support label switching an ATM switch must implement the control component of label switching. This consists primarily of label allocation and maintenance procedures. Label binding information is communicated by several mechanisms, notably the Label Distribution Protocol (LDP). Candidate protocols being considered by the LDP design team are described in [2, 4]. This document imposes certain requirements on the LDP. Since the label switching control component uses information learned directly from network layer routing protocols, this implies that the switch must participate as a peer in these protocols (e.g., OSPF, IS-IS). In some cases, LSRs make use of other protocols (e.g. RSVP, PIM, BGP) to distribute label bindings. In these cases, an ATM LSR would need to participate in these protocols. Support of label switching on an ATM switch does not require the switch to support the ATM control component defined by the ITU and ATM Forum (e.g., UNI, PNNI). An ATM-LSR may optionally respond to OAM cells. Davie, et al. [Page 4] Internet Draft draft-davie-mpls-atm-00.txt October 1997 5. Hybrid Switches (Ships in the Night) The existence of the label switching control component on an ATM switch does not preclude the ability to support the ATM control component defined by the ITU and ATM Forum on the same switch and the same interfaces. The two control components, label switching and the ITU/ATM Forum defined, would operate independently. Definition of how such a device operates is beyond the scope of this document. However, only a small amount of information needs to be consistent between the two control components, such as the portions of the VPI/VCI space which are available to each component. 6. Use of VPI/VCIs Label switching is accomplished by associating labels with routes and using the label value to forward packets, including determining the value of any replacement label. See [1] for further details. In an ATM-LSR, the label is carried in the VPI and/or VCI field. Just as in conventional ATM, for a cell arriving at an interface, the VPI/VCI is looked up, replaced, and the cell is switched. ATM-LSRs may be connected by ATM virtual paths to enable interconnection of ATM-LSRs over a cloud of conventional ATM switches. In this case, the label is carried in the VCI field. For two connected ATM-LSRs, a connection must be available for LDP. The default is for this connection to be on VPI 0, VCI 32. For ATM- LSRs connected by a VPI of value x, the default for the LDP connection is VPI x, VCI 32. Additionally, for all VPI values, VCIs 0 - 32 are not used as labels. With the exception of these reserved values, the VPI/VCI values used in the two directions of the link may be treated as independent spaces. The allowable ranges of VPI/VCIs are always communicated through LDP. If more than one VPI is used for label switching, the allowable range of VCIs may be different for each VPI, and each range is communicated through LDP. Davie, et al. [Page 5] Internet Draft draft-davie-mpls-atm-00.txt October 1997 7. Label Allocation and Maintenance Procedures ATM-LSRs use the downstream-on-demand allocation mechanism described in [1]. The procedures for label allocation depend on whether the switches support VC-merge or not. We therefore describe the two scenarios in turn. We begin by describing the behavior of members of the Edge Set of an ATM-LSR cloud; these edge LSRs are not themselves ATM-LSRs, and their behavior is the same whether the cloud contains VC-merge capable LSRs or not. 7.1. Edge LSR Behavior Consider a member of the Edge Set of an ATM-LSR cloud. Assume that, as a result of its routing calculations, it selects an ATM-LSR as the next hop of a certain route, and that the next hop is reachable via a LC-ATM interface. The Edge LSR uses LDP's binding request message to request a label binding from the next hop. The hop count field in the request is set to 1. Once the Edge LSR receives the label binding information, the label is used as an outgoing label. The binding received by the edge LSR may contain a hop count, which represents the number of hops a packet will take to cross the ATM-LSR cloud when using this label. The edge LSR may either - use this hop count to decrement the TTL of packets before transmitting them over the cloud - decrement the TTL of packets by one before transmitting them over the cloud. The choice between these two options should be made based on local configuration. When a member of the Edge Set of the ATM-LSR cloud receives a label binding request from an ATM-LSR, it allocates a label, creates a new entry in its Label Information Base (LIB), places that label in the incoming label component of the entry, and returns (via LDP) a binding containing the allocated label back to the peer that originated the request. It sets the hop count in the binding to 1. When a routing calculation causes an Edge LSR to change the next hop for a route, and the former next hop was in the ATM-LSR cloud, the Edge LSR should notify the former next hop (via LDP) that the label binding associated with the route is no longer needed. Davie, et al. [Page 6] Internet Draft draft-davie-mpls-atm-00.txt October 1997 7.2. Conventional ATM Switches (non-VC-merge) When an ATM-LSR receives (via LDP) a label binding request for a certain route from a peer connected to the ATM-LSR over a LC-ATM interface, the ATM-LSR takes the following actions: - it allocates a label, creates a new entry in its Label Information Base (LIB), and places that label in the incoming label component of the entry; - it requests (via LDP) a label binding from the next hop for that route; - it returns (via LDP) a binding containing the allocated incoming label back to the peer that originated the request. The hop count field in the request that the ATM-LSR sends (to the next hop LSR) is set to the hop count field in the request that it received from the upstream LSR plus one. Once the ATM-LSR receives the binding from the next hop, it places the label from the binding into the outgoing label component of the LIB entry. The ATM-LSR may choose to wait for the request to be satisfied from downstream before returning the binding upstream (a "conservative" approach). In this case, the ATM-LSR increments the hop count it received from downstream and uses this value in the binding it returns upstream. If the value of the hop count equals MAX_HOP_COUNT the ATM-LSR should notify the upstream neighbor that it could not satisfy the binding request. Alternatively, the ATM-LSR may return the binding upstream without waiting for a binding from downstream (an "optimistic" approach). In this case, it uses a reserved value for hop count in the binding, indicating that it is unknown. The correct value for hop count will be returned later, as described below. Since both the conservative and the optimistic approaches have advantages and disadvantages, this is left as an implementation choice. Note that an ATM-LSR, or a member of the edge set of an ATM-LSR cloud, may receive multiple binding requests for the same route from the same ATM-LSR. It must generate a new binding for each request (assuming adequate resources to do so), and retain any existing binding(s). For each request received, an ATM-LSR should also generate a new binding request toward the next hop for the route. Davie, et al. [Page 7] Internet Draft draft-davie-mpls-atm-00.txt October 1997 When a routing calculation causes an ATM-LSR to change the next hop for a route, the ATM-LSR should notify the former next hop (via LDP) that the label binding associated with the route is no longer needed. When a LSR receives a notification that a particular label binding is no longer needed, the LSR may deallocate the label associated with the binding, and destroy the binding. In the case where an ATM-LSR receives such notification and destroys the binding, it should notify the next hop for the route that the label binding is no longer needed. If a LSR does not destroy the binding, it may re-use the binding only if it receives a request for the same route with the same hop count as the request that originally caused the binding to be created. When a route changes, the label bindings are re-established from the point where the route diverges from the previous route. LSRs upstream of that point are (with one exception, noted below) oblivious to the change. Whenever a LSR changes its next hop for a particular route, if the new next hop is an ATM-LSR or a member of the edge set reachable via a LC-ATM interface, then for each entry in its LIB associated with the route the LSR should request (via LDP) a binding from the new next hop. When an ATM-LSR receives a label binding from a downstream neighbor, it may already have provided a corresponding label binding for this route to an upstream neighbor, either because it is operating optimistically or because the new binding from downstream is the result of a routing change. In this case, it should extract the hop count from the new binding and increment it by one. If the new hop count is different from that which was previously conveyed to the upstream neighbor (including the case where the upstream neighbor was given the value `unknown') the ATM-LSR must notify the upstream neighbor of the change. Each ATM-LSR in turn increments the hop count and passes it upstream until it reaches the ingress Edge LSR. If at any point the value of the hop count equals MAX_HOP_COUNT, the ATM- LSR should withdraw the binding from the upstream neighbor. Whenever an ATM-LSR originates a label binding request to its next hop LSR as a result of receiving a label binding request from another (upstream) LSR, and the request to the next hop LSR is not satisfied, the ATM-LSR should destroy the binding created in response to the received request, and notify the requester (via LDP). If an ATM-LSR receives a binding request containing a hop count that equals MAX_HOP_COUNT, no binding should be established and an error message should be returned to the requester. Davie, et al. [Page 8] Internet Draft draft-davie-mpls-atm-00.txt October 1997 When a LSR determines that it has lost its LDP session with another LSR, the following actions are taken. Any binding information learned via this connection must be discarded. For any label bindings that were created as a result of receiving label binding requests from the peer, the LSR may destroy these bindings (and deallocate labels associated with these binding). An ATM-LSR should use `split-horizon' when it satisfies binding requests from its neighbors. That is, if it receives a request for a binding to a particular route and the LSR making that request is, according to this ATM-LSR, the next hop for that route, it should not return a binding for that route. Note that it is not possible for complete label-switched looping paths to be established when VC merge is not supported. The worst case behavior (possible only if optimistic mode is used) is that binding request messages could travel in a loop, setting up a label switched spiral, which would then be torn down when the hop count in the binding request reached MAX_HOP_COUNT. In conservative mode, no label switched path can be set up unless the path exits the ATM-LSR cloud. 7.3. VC-merge-capable ATM Switches Relatively minor changes are needed to accommodate ATM-LSRs which support VC-merge. The primary difference is that a VC-merge-capable ATM-LSR needs only one outgoing label per route, even if multiple requests for label bindings to that route are received from upstream neighbors. When a VC-merge-capable ATM-LSR receives a binding request from an upstream LSR for a certain route, and it does not already have an outgoing label binding for that route, it issues a bind request to its next hop just as before. If, however, it already has an outgoing label binding for that route, it does not need to issue a downstream binding request. Instead, it creates a new LIB entry, allocates an incoming label for that entry and returns that label in a binding to the upstream requester, and uses the existing outgoing label for the outgoing label entry in the LIB. It also takes the hop count that was provided with the label binding it received from downstream, increments it by one, and uses this value in the binding that it sends to the upstream requester. Note that, just like conventional ATM-LSRs and members of the edge set of the ATM-LSR cloud, a VC-merge-capable ATM-LSR must issue a new binding every time it receives a request from upstream, since there may be switches upstream which do not support VC-merge. However, it Davie, et al. [Page 9] Internet Draft draft-davie-mpls-atm-00.txt October 1997 only needs to issue a corresponding binding request downstream if it does not already have a label binding for the appropriate route. When a change in the routing table of a VC-merge-capable ATM-LSR causes it to select a new next hop for one of its routes, it releases the binding for that route from the former next hop and requests a new binding from the new next hop. If the new binding contains a hop count that differs from that which was received in the old binding, then the ATM-LSR must take the new hop count, increment it by one, and notify any upstream neighbors who have label bindings for this route of the new value. Just as with conventional ATM-LSRs, this enables the new hop count to propagate back towards the ingress of the ATM-LSR cloud. If at any point the hop count reaches MAX_HOP_COUNT, then the label bindings for this route must be withdrawn from all upstream neighbors to whom a binding was previously provided. This ensures that any loops caused by routing transients will be detected and broken. While the choice between "conservative" and "optimistic" binding remains for VC-merge capable LSRs, the advantages of the conservative approach are greater in this case. This is because: - the optimistic mode permits the formation of a looping switched path which will not be removed until routing changes to remove the loop; - the conservative mode does not have the same drawbacks when VC merge is supported, since it will often be possible to obtain a binding by merging into an existing path, without waiting for binding requests and responses to propagate across the entire ATM-LSR cloud. The use of conservative mode along with hop counts in the binding requests and responses ensures that stable looping paths cannot be set up. Implementations may choose to support the diffusion algorithm described in [1] for stronger protection against transient loops. Davie, et al. [Page 10] Internet Draft draft-davie-mpls-atm-00.txt October 1997 7.4. Efficient use of label space The above discussion assumes that an edge LSR will request one label for each prefix in its routing table that has a next hop in the ATM- LSR cloud. In fact, it is possible to significantly reduce the number of labels needed by having the edge LSR request instead one label for several routes. Use of many-to-one mappings between routes (address prefixes) and labels using the notion of Forwarding Equivalence Classes (as described in [1]) provides a mechanism to conserve the number of labels. 8. Encapsulation By default, all labelled packets should be transmitted with the generic label encapsulation, as defined in [3]. A packet thus encapsulated is then directly encapsulated within an AAL5 frame. Since the value at the top of the label stack is determined from the VCI and/or VPI fields, the top label in the label encapsulation is ignored. Other fields in the top stack entry, such as TTL and COS, may be used by LSRs at the edges of the ATM-LSR cloud that are able to examine that entry. For systems which are using only one level of labelling, it is theoretically possible to use a null encapsulation. In this case, IP packets are carried directly inside AAL5 frames, as in the null encapsulation of RFC 1483. However, all ATM-LSRs in a cloud must be configured to use this encapsulation to avoid reassembly and re- encapsulation in the middle of the cloud. The initial LDP connection, described in Section 5, uses the LLC/SNAP encapsulation, as defined in Section 4.1 of RFC1483. This same VCI (with the LLC/SNAP encapsulation) may be used to exchange Network Layer routing information as well. LDP may be used to advertise additional VPI/VCIs to carry control information or non-labelled packets. These may use either the null encapsulation, as defined in Section 5.1 of RFC1483, or the LLC/SNAP encapsulation, as defined in Section 4.1 of RFC1483. Davie, et al. [Page 11] Internet Draft draft-davie-mpls-atm-00.txt October 1997 9. Security Considerations Security considerations are not addressed in this document. 10. Intellectual Property Considerations Cisco Systems may seek patent or other intellectual property protection for some or all of the technologies disclosed in this document. If any standards arising from this document are or become protected by one or more patents assigned to Cisco Systems, Cisco intends to disclose those patents and license them under openly specified and non-discriminatory terms, for no fee. 11. References [1] Rosen, E. et al. A Proposed Architecture for MPLS, Internet Draft, draft-ietf-mpls-arch-00.txt, August, 1997 [2] Doolan, P. et al. Tag Distribution Protocol, Internet Draft, draft-doolan-tdp-spec-01.txt, May, 1997. [3] Rosen, E. et al. Label Switching: Label Stack Encodings, Internet Draft, draft-rosen-tag-stack-03.txt, July, 1997. [4] Feldman, N., and Viswanathan, A. ARIS Specification, Internet Draft, draft-feldman-aris-spec-00.txt, March, 1997. 12. Acknowledgments Significant contributions to this work have been made by Anthony Alles, Fred Baker, Dino Farinacci, Guy Fedorkow, Arthur Lin, Morgan Littlewood and Dan Tappan. 13. Authors' Addresses Bruce Davie Cisco Systems, Inc. 250 Apollo Drive Chelmsford, MA, 01824 E-mail: bsd@cisco.com Davie, et al. [Page 12] Internet Draft draft-davie-mpls-atm-00.txt October 1997 Paul Doolan Ennovate Networks Inc. 330 Codman Hill Rd Boxborough, MA 01719 E-mail: pdoolan@ennovatenetworks.com Jeremy Lawrence Cisco Systems, Inc. 1400 Parkmoor Ave. San Jose, CA, 95126 E-mail: jlawrenc@cisco.com Keith McCloghrie Cisco Systems, Inc. 170 Tasman Drive San Jose, CA, 95134 E-mail: kzm@cisco.com Yakov Rekhter Cisco Systems, Inc. 170 Tasman Drive San Jose, CA, 95134 E-mail: yakov@cisco.com Eric Rosen Cisco Systems, Inc. 250 Apollo Drive Chelmsford, MA, 01824 E-mail: erosen@cisco.com George Swallow Cisco Systems, Inc. 250 Apollo Drive Chelmsford, MA, 01824 E-mail: swallow@cisco.com Davie, et al. [Page 13]