[Docs] [txt|pdf] [Tracker] [Email] [Nits]

Versions: 00 01 02 03 04 05 06 07 RFC 4973

Network Working Group                                       P. Srisuresh
INTERNET-DRAFT                                                 P. Joseph
Expires as of December 25, 2001                         Jasmine Networks
                                                              June, 2001

            New TE LSAs to extend OSPF 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.

   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-

   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

   The list of Internet-Draft Shadow Directories can be accessed at


   OSPF is a well established link state routing protocol used for
   topology discovery and computing forwarding table based on
   shortest-Path criteria. Traffic Engineering extensions (OSPF-TE)
   will use criteria different from shortest-path so as to route
   traffic around congestion paths and meet varying Service Level
   agreements. OSPF-TE may also be used by non-IP networks such as
   photonic and TDM (SONET/SDH) circuit switch networks for
   light-path or TDM circuit setup between two end-points. The
   approach outlined in this document differs from that of
   [OPQLSA-TE]. The document does not suggest the use of Opaque LSAs
   to add TE extensions to OSPF. Rather, new TE LSAs, modeled after
   existing LSAs and flooding scope are proposed to overcome the
   scaling limitations of the approach outlined in [OPQLSA-TE]. The
   document draws a distinction between TE and non-TE topologies and

Srisuresh & Joseph                                              [Page 1]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   restricts flooding of TE LSAs into non-TE topology.  The document
   covers OSPF extensions for packet and non-packet networks alike,
   providing a unified extension mechanism for all networks. As such,
   this approach improves interoperability between peer network
   elements. Lastly, the document specifies a transition path for
   vendors currently using opaque LSAs to transition to using new
   TE LSAs outlined here.

1. Introduction

   There is substantial industry experience with deploying OSPF link
   state routing protocol. That makes OSPF a good candidate to adapt for
   traffic engineering purposes. The dynamic discovery of network
   topology, flooding algorithm and the hierarchical organization of
   areas can all be used effectively in creating and tearing traffic
   links on demand. The intent of the document is to build an abstract
   view of the topology in conjunction with the traffic engineering
   characteristics of the nodes and links involved.

   The connectivity topology may remain relatively stable, except when
   the existing links or networking nodes go down or flap or new nodes
   and links are added to the network. The objective of traffic
   engineering is to set up circuit path(s) across a pair of nodes or
   links, as the case may be, so as to forward traffic of a certain
   forwarding equivalency class. Circuit emulation in a packet network
   is accomplished by each MPLS intermediary node performing label
   swapping. Whereas, circuit emulation in a TDM or Fiber cross-connect
   network is accomplished by configuring the switch fabric in each
   intermediary node to do the appropriate switching (TDM, fiber or
   Lamda) for the duration of the circuit.

   The objective of this document is not how to set up traffic circuits,
   but rather provide the necessary TE parameters for the nodes and
   links that constitute the TE topology. Unlike the traditional OSPF,
   the TE extended OSPF will be used to build circuit paths, meeting
   certain TE criteria. The only requirement is that end-nodes and/or
   end-links of a circuit be identifiable with an IP address. For
   non-IP networks (such as TDM or photonic cross connect networks),
   Mapping IP addresses to a well-known name can be done through a
   DNS-like mechanism.

   The approach suggested in this document is different from the
   Opaque-LSA-based approach outlined in [OPQLSA-TE]. Section 10
   compares the two approaches and outlines a strategy to transition
   from Opaque-LSA based deployments to the new-TE-LSA approach
   outlined here.

Srisuresh & Joseph                                              [Page 2]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

2. Traffic Engineering

   A traffic engineered circuit may be identified by the tuple of
   (Forwarding Equivalency Class, TE parameters for the circuit,
   Origin Node/Link, Destination node/Link).

   The forwarding Equivalency class(FEC) may be constituted of a number
   of criteria such as (a) Traffic arriving on a specific interface,
   (b) Traffic meeting a certain classification criteria (ex: based on
   fields in the IP and transport headers), (c) Traffic in a certain
   priority class, (d) Traffic arriving on a specific set of TDM (STS)
   circuits on an interface, (e) Traffic arriving on a certain
   wave-length of an interface, (f) Traffic arriving at a certain time
   of day, and so on. A FEC may be constituted as a combination of one
   or more of the above criteria. Discerning traffic based on the FEC
   criteria is a mandatory requirement on Label Edge Routers (LERs).
   Traffic content is transparent to the Intermediate Label Switched
   Routers (LSRs), once a circuit is formed. LSRs are simply
   responsible for keeping the circuit in-tact for the lifetime of the
   circuit(s). As such, this document will not address FEC or the
   associated signaling to setup circuits. [MPLS-TE] and [GMPLS-TE]
   address the FEC criteria. Whereas, [RSVP-TE] and [CR-LDP] address
   different types of signaling protocols.

   As for TE parameters for the circuit, this refers to the TE
   parameters for all the nodes and links constituting a circuit.
   Typically, TE parameters for a node in a TE circuit may include
   the following.

        * Traffic prioritization ability,
        * Ability to provision bandwidth on interfaces,
        * Support of CSPF algorithms,
        * TE-Circuit switch type,
        * Automatic protection switching.

   TE parameters for the link include:

        * Bandwidth availability,
        * reliability of the link,
        * color assigned to the link
        * cost of bandwidth usage on the link.
        * membership to a Shared Risk Link Group and so on.

   Only the unicast paths circuit paths are considered here. Multicast
   variations are currently considered out of scope for this document.
   The requirement is that the originating as well as the terminating
   entities of a TE path are identifiable by their IP address.

Srisuresh & Joseph                                              [Page 3]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

3. Terminology

   Definitions for majority of the terms used in this document with
   regard to OSPF protocol may be found in [OSPF-V2]. MPLS and traffic
   engineering terms may be found in [MPLS-ARCH]. RSVP-TE and CR-LDP
   signaling specific terms may be found in [RSVP-TE] and [CR-LDP]

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALLNOT",
   this document are to be interpreted as described in RFC 2119.

   Below are definitions for the terms used within this document.

3.1. OSPF-TE router (or) TE-compliant OSPF router

   This is a router that supports the OSPF TE extensions described
   in this document. This requires that at least one of the links
   attached to the router support TE extensions and, at least one of
   the links attached to the router support Packet termination and
   run the OSPF-TE protocol.

   An OSPF-TE router supports native OSPF as well as the TE
   extensions outlined here.

3.2. Native OSPF router

   A native OSPF router is an OSPF router that does not support
   the TE extensions described in this document. An autonomous
   system could be constituted of a combination of native OSPF
   routers and OSPF-TE routers.

   A native OSPF router, when enhanced to include the extensions
   described in this document can become a OSPF-TE router.

3.3. TE nodes vs. non-TE (bormal) nodes

   A TE-Node is an intermediate or edge node taking part in the
   traffic engineered (TE) network. Specifically, a TE circuit
   is constituted of a series of TE nodes connected to each other
   via the TE links.

   A non-TE node or a normal node is a node that does not have any
   TE links attached to it and does not take part in a TE network.
   Specifically, native OSPF-nodes that do not take part in a TE
   network fall under this category.

Srisuresh & Joseph                                              [Page 4]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

3.4. TE links vs. non-TE (normal) links

   A TE Link is a network attachment that supports traffic
   engineering. Specifically, a TE circuit can only be setup using
   a combination of TE nodes and TE links connected to each other.

   Non-TE links or a normal link is one that that does not
   support traffic engineering. For example, native OSPF protocol
   and least cost criteria may be used to determine reachability
   of subnets in a network constituted of normal OSPF nodes and
   normal OSPF links.

3.5. Packet interface vs. non-packet interface

   Interfaces on an OSPF-TE router may be characterized as those
   that can terminate (i.e., send and receive) IP packet data and
   those that can not. Both types of links can be part of a
   traffic engineered network. In contrast, a native OSPF router
   does not support non-packet interfaces.

   Needless to say, the OSPF protocol and its TE extensions can only
   be enabled on interfaces supporting IP packet termination.

3.6. TE topology vs. non-TE topology

   A TE topology is constituted of a set of contiguous TE nodes and
   TE links. Associated with each TE node and TE link is a set of TE
   criteria that may be supported at any given time. A TE topology
   allows circuits to be overlayed statically or dynamically based
   on a specific TE criteria.

   A non-TE topology specifically refers to the network that does not
   support TE. Control protocols such as OSPF may be run on the non-TE
   topology. IP forwarding table used to forward IP packets on this
   network is built based on the control protocol specific algorithm,
   such as OSPF shortest-path criteria.

3.7. TLV

   A TLV, strictly stands for an object in the form of Tag-Length-Value.
   However, this term is also used in the document, at times, to simply
   refer a Traffic Engineering attribute of a TE-node or TE-link.

   All TLVs are assumed to be of the following format, unless specified
   otherwise. The Tag and length are 16 bits wide each. The length
   includes the 4 bytes required for Tag and Length specification.

Srisuresh & Joseph                                              [Page 5]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

        0                   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
       |            Tag                |     Length (4 or more)        |
       |                            Value ....                         |
       |                            ....                               |

3.8. Router-TE TLVs

     TLVs used to describe the TE capabilities of a TE-node.

3.9. Link-TE TLVs

     TLVs used to describe the TE capabilities of a TE-link.

4. Motivation and Implicit assumptions for the TE extensions

   The motivation behind the OSPF-TE described in this document is to
   dynamically discover the TE-network topology, devise a scalable
   flooding methodology and benefit from the hierarchical area
   organization and other techniques of the native OSPF. The result
   would be the ability to build an abstract view of a network
   topology with all the traffic engineering characteristics.

   With traditional OSPF, the goal is to build a forwarding table to
   reach various subnets in the IP network with least-cost as the
   basis. However, the goal of OSPF-TE is to determine a circuit path
   (that can be pinned-down for a desired duration) meeting a certain
   set of traffic engineering criteria. Further, the circuit path
   could consist entirely of nodes and links that do not carry IP

   The following assumptions are made throughout the document for
   the discussion of OSPF-TE.

   1. Interfaces on an OSPF-TE router may be characterized as those
      that can terminate (i.e., send and receive) IP packet data and
      those that wont. Both types of links can be part of a traffic
      engineered network. Needless to say, the OSPF-TE protocol can
      only be enabled on interfaces that support IP packet data
      termination. And, TE LSAs may be exchanged over non-TE links.

   2. Unlike traditional OSPF, OSPF-TE protocol must be capable of

Srisuresh & Joseph                                              [Page 6]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

      advertising link state of interfaces that are not capable of
      handling packet data. As such, the OSPF-TE protocol cannot be
      required to synchronize its link-state database with neighbors
      across all its links. It is sufficient to synchronize
      link-state database in an area, across a subset of the links -
      say, the packet terminating interfaces. Yet, the TE LSDB
      (LSA database) should be synchronized across all OSPF-TE nodes
      within an area.

      All interfaces or links described by the TE LSAs will be
      present in the TE topology database (a.k.a. TE LSDB).

   3. An OSPF-TE router with links in an OSPF area will need to
      establish router adjacency with at least one other OSPF-TE
      neighboring router in order for the router's database to be
      synchronized with other routers in the area. Failing this, the
      OSPF router will not be in the TE calculations of other TE
      routers in the area. Refer [OSPF-FL1] for flooding

      However, two routers that are physically connected to the same
      link (or broadcast network) neednt be router adjacent via the
      Hello protocol, if the link is not packet terminated.

   4. Each IP subnet on a TE-configurable network MUST have a minimum
      of one node with an interface running OSPF-TE protocol. This is
      despite the fact that all nodes on the subnet may take part in
      Traffic Engineering. (Example: SONET/SDH TDM ring with a single
      Gateway Network Element, a.k.a. GNE running the OSPF protocol,
      yet all other nodes in the ring are also full members of a TE

      An OSPF-TE node may advertise more than itself and the links
      it is directly attached to. It may also advertise other TE
      participants and their links, on their behalf.

   5. As a general rule, all nodes and links that may be party
      to a TE circuit should be uniquely  identifiable by an IP
      address. As for router ID, a separate loopback IP address
      for the router, independent of the links attached, is

   6. TE nodes may have 2 types of link state databases - a normal
      LSDB and a TE-LSDB. A normal LSDB, constituted of non-TE
      links and nodes attached to these links(i.e., non-TE as well
      as TE nodes), will use shortest-path criteria to forward IP
      packets over normal non-TE links. The TE-LSDB, constituted
      of TE nodes and TE links, may be used to setup TE circuit

Srisuresh & Joseph                                              [Page 7]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

      paths along the TE topology.

5. The OSPF Options field

   A new TE flag is introduced to identify TE extensions to the OSPF.
   With this, the OSPF v2 will have no more reserved bits left for
   future option extensions. This bit will be used to distinguish
   between routers that support Traffic engineering extensions and
   those that do not.

   The OSPF options field is present in OSPF Hello packets, Database
   Description packets and all link state advertisements.  See
   [OSPF-V2], [OSPF-NSSA] and [OPAQUE] for a description of the
   bits in options field. Only the TE-Bit is described in this

                   |TE | O | DC | EA | N/P | MC | E | * |

                       The OSPF options field - TE support

   TE-Bit: This bit is set to indicate support for Traffic Engineering
           extensions to the OSPF. The Hello protocol which is used for
           establishing router adjacency and bidirectionality of the
           link will use the TE-bit to build adjacencies between two
           nodes that are either both TE-compliant or not. Two routers
           will not become TE-neighbors unless they agree on the state
           of the TE-bit. TE-compliant OSPF extensions are advertised
           only to the TE-compliant routers. All other routers may
           simply ignore the advertisements.

6. Bringing up TE adjacencies; TE vs. Non-TE topologies

   OSPF creates adjacencies between neighboring routers for the purpose
   of exchanging routing information. In the following subsections, we
   describe the use of Hello protocol to establish TE capability
   compliance between neighboring routers of an area. Further, the
   capability is used as the basis to build a TE vs. non-TE network

6.1.  The Hello Protocol

   The Hello Protocol is primarily responsible for dynamically
   establishing and maintaining neighbor adjacencies. In a TE network,
   it may not be required or possible for all links and neighbors to
   establish adjacency using this protocol.

Srisuresh & Joseph                                              [Page 8]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   The Hello protocol will use the TE-bit to establish Traffic
   Engineering capability (or not) between two OSPF routers.

   For NBMA and broadcast networks, this protocol is responsible for
   electing the designated router and the backup designated router.
   For a TDM ring network, the designated and backup designated
   routers may either be preselected (ex: GNE, backup-GNE) or
   determined via the same Hello protocol. In any case, routers
   supporting the TE option shall be given a higher precedence
   for becoming a designated router over those that donot support TE.

6.2. Flooding and the Synchronization of Databases

   In OSPF, adjacent routers within an area must synchronize their
   databases.   However, as observed in [OSPF-FL1], the requirement
   may be stated more concisely that all routers in an area must
   converge on the same link state database. To do that, it suffices
   to send single copies of LSAs to the neighboring routers in an
   area, rather than send one copy on each of the connected
   interfaces. [OSPF-FL1] describes in detail how to minimize
   flooding (Initial LSDB synchronization as well as the
   asynchronous LSA updates) within an area.

   With the OSPF-TE described here, a TE-only network topology is
   constructed based on the TE option flag in the Hello packet.
   Subsequent to that, TE LSA flooding in an area is limited to
   TE-only routers in the area, and do not impact non-TE routers
   in the area. A network may be constituted of a combination of
   a TE topology and a non-TE (control) topology. Standard IP
   packet forwarding and routing protocols are possible along the
   control topology.

   In the case where some of the neighbors are TE compliant and
   others are not, the designated router will exchange different
   sets of LSAs with its neighbors. TE LSAs are exchanged only
   with the TE neighbors. Normal LSAs do not include description
   for TE links. As such, normal LSAs are exchanged with all
   neighbors (TE and non-TE alike) over a shared non-TE link.

   Flooding optimization in a TE network is essential
   for two reasons. First, the control traffic for a TE network is
   likely to be much higher than that of a non-TE network. Flooding
   optimizations help to minimize the announcements and the
   associated retransmissions and acknowledgements on the network.
   Secondly, the TE nodes need to converge at the earliest to keep
   up with TE state changes occuring throughout the TE network.

Srisuresh & Joseph                                              [Page 9]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   This process of flooding along a TE topology cannot be folded
   into the Opaque-LSA based TE scheme ([OPQLSA-TE]), because
   Opaque LSAs (say, LSA #10) have a pre-determined flooding
   scope. Even as a TE topology is available from the use of
   TE option flag, the TE topology is not usable for flooding
   unless a new TE LSA is devised, whose boundaries can be set to
   span the TE-only routers in an area.

   NOTE, a new All-SPF-TE Multicast address may be used for the
   exchange of TE compliant database descriptors.

6.3.  The Designated Router

   The Designated Router is elected by the Hello Protocol on broadcast
   and NBMA networks. In general, when a router's interface to a
   network first becomes functional, it checks to see whether there is
   currently a Designated Router for the network.  If there is, it
   accepts that Designated Router, regardless of its Router Priority,
   so long as the current designated router is TE compliant. Otherwise,
   the router itself becomes Designated Router if it has the highest
   Router Priority on the network and is TE compliant.

   Clearly, TE-compliance must be implemented on the most robust
   routers only, as they become most likely candidates to take on
   additional role as designated router.

   Alternatively, there can be two sets of designated routers, one for
   the TE compliant routers and another for the native OSPF routers
   (non-TE compliant).

6.4.  The Backup Designated Router

   The Backup Designated Router is also elected by the Hello
   Protocol.  Each Hello Packet has a field that specifies the
   Backup Designated Router for the network. Once again, TE-compliance
   must be weighed in conjunction with router priority in determining
   the backup designated router. Alternatively, there can be two sets
   of backup designated routers, one for the TE compliant routers and
   another for the native OSPF routers (non-TE compliant).

6.5.  The graph of adjacencies

   An adjacency is bound to the network that the two routers have
   in common.  If two routers have multiple networks in common,
   they may have multiple adjacencies between them. The adjacency
   may be split into two different types - Adjacency between
   TE-compliant routers and adjacency between non-TE compliant
   routers. A router may choose to support one or both types of

Srisuresh & Joseph                                             [Page 10]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001


   Two graphs are possible, depending on whether a Designated
   Router is elected for the network.  On physical point-to-point
   networks, Point-to-MultiPoint networks and virtual links,
   neighboring routers become adjacent whenever they can
   communicate directly.  The adjacency can only be one of
   (a) TE-compliant or (b) non-TE compliant. In contrast, on
   broadcast and NBMA networks the Designated Router and the
   Backup Designated Router may maintain two sets of adjacency.
   However, the remaining routers will participate in either
   TE-compliant adjacency or non-TE-compliant adjacency, but not
   both.  In the Broadcast network below, you will notice that
   routers RT7 and RT3 are chosen as the designated and backup
   routers respectively. Within the network, Routers RT3, RT4
   and RT7 are TE-compliant. RT5 and RT6 are not. So, you will
   notice the adjacency variation with RT4 vs. RT5 or RT6.

          +---+            +---+
          |RT1|------------|RT2|            o---------------o
          +---+    N1      +---+           RT1             RT2

            +---+   +---+   +---+                /|:        :
            |RT7|   |RT3|   |RT4|               / | :       :
            +---+   +---+   +---+              /  |  :      :
              |       |       |               /   |   :     :
         +-----------------------+        RT5o RT6o    oRT4 :
                  |       |     N2            *   *   ;     :
                +---+   +---+                  *  *  ;      :
                |RT5|   |RT6|                   * * ;       :
                +---+   +---+                    **;        :

       Figure 6: The graph of adjacencies with TE-compliant routers.

7. New TE LSAs

   The native OSPF protocol has a total of 11 LSA types. Definitions
   for LSA types 1 through 5 may be found in [OSPF-v2]. LSA type 6 is

Srisuresh & Joseph                                             [Page 11]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   defined in [MOSPF]. LSA type 7 definition may be found in [NSSA].
   LSA type 8 may be found in  [BGP-OSPF]. Lastly, LSA types 9 through
   11 are defined in [OPAQUE].

   Each of the LSA types defined are different in content and flooding
   scope. For instance, Opaque LSA types 9 through 11 are general
   purpose LSAs, with flooding scope set to link-local, area-local and
   AS-wide (except into stub areas) respectively. As will become
   apparent soon, the boundaries for Opaque LSAs are not appropriate
   for flooding TE data.

   In the following subsections, we define new LSAs for Traffic
   engineering use. The new TE LSAs are largely modeled after the
   existing LSAs for content format and flooding scope. The LSA types
   are assigned such that the high bit of the LS-type octet is set
   to 1. Standard link-state database flooding mechanisms (with
   optimizations discussed in previous sections) are used for
   distribution of TE LSAs along the  TE-restricted topology. The
   flooding scope is also defined for each of the newly defined TE

7.1. TE-Router LSA

   Router LSAs are Type 1 LSAs. The TE-router LSA is modeled after the
   router LSA with the same flooding scope as the router-LSA, except
   that the scope is further restricted to TE-only nodes within the
   area. A value of 0x81 is assigned to TE-router LSA. The TE-router
   LSA describes the router-TE metrics as well as the link-TE metrics
   of the TE links attached to the router. Below is the format of the
   TE-router LSA. Unless specified explicitly otherwise, the fields
   carry the same meaning as they do in a router LSA. Only the
   differences are explained below.

        0                   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
       |            LS age             |     Options   |     0x81      |
       |                        Link State ID                          |
       |                     Advertising Router                        |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |    0    |V|E|B|      0        |       Router-TE flags         |

Srisuresh & Joseph                                             [Page 12]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

       |  Router-TE flags (contd.)     |       Router-TE TLVs          |
       |                     ....                                      |
       |                     ....      |            # of TE links      |
       |                          Link ID                              |
       |                         Link Data                             |
       |     Type      |        0      |    Link-TE flags              |
       |   Link-TE flags (contd.)      |  Zero or more Link-TE TLVs    |
       |                          Link ID                              |
       |                         Link Data                             |
       |                              ...                              |

        In TE-capable router nodes, the TE-compliance bit is set to 1.

   Router-TE flags field (TE capabilities of the router node)

        Below is an initial set of definitions. More may be standardized
        if necessary. The TLVs are not expanded in the current rev. Will
        be done in the follow-on revs. The field imposes a restriction
        of no more than 32 flags to describe the TE capabilities of a

       |L|L|P|T|L|F|                                           |S|S|S|C|
       |S|E|S|D|S|S|                                           |T|E|I|S|
       |R|R|C|M|C|C|                                           |A|L|G|P|
       |<---- Boolean TE flags ------->|<- TE flags pointing to TLVs ->|

       Bit LSR
           When set, the router is considered to have LSR capability.

       Bit LER
           When set, the router is considered to have LER capability.
           All MPLS border routers will be required to have the LER

Srisuresh & Joseph                                             [Page 13]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

           capability. When the E bit is also set, that indicates an
           AS Boundary router with LER capability. When the B bit is
           also set, that indicates an area border router with LER

       Bit PSC
           Indicates the node is Packet Switch Capable.

       Bit TDM
           Indicates the node is TDM circuit switch capable.

       Bit LSC
           Indicates the node is Lamda switch Capable.

       Bit FSC
           Indicates the node is Fibre (can also be a non-fibre link
           type) switch capable.

       Bit STA
           Label Stack Depth limit TLV follows. This is applicable only
           when the PSC flag is set.

       Bit SEL
           TE Selection Criteria TLV, supported by the router, follows.

       Bit SIG
           MPLS Signaling protocol support TLV follows.

       BIT CSPF
           CSPF algorithm support TLV follows.

   Router-TE TLVs
      The following Router-TE TLVs are defined.

      TE-selection-Criteria TLV (Tag ID = 1)

          The values can be a series of resources  that may be used
          as the criteria for traffic engineering (typically with the
          aid of a signaling protocol such as RSVP-TE or CR-LDP or LDP).

                    - Bandwidth based LSPs (1)
                    - Priority based LSPs  (2)
                    - Backup LSP           (3)
                    - Link cost            (4)

          Bandwidth criteria is often used in conjunction with Packet
          Switch Capable nodes. The unit of bandwidth permitted to be

Srisuresh & Joseph                                             [Page 14]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

          configured may however vary from vendor to vendor. Bandwidth
          criteria may also be used in conjunction with TDM nodes. Once
          again, the granularity of bandwidth allocation may vary from
          vendor to vendor.

          Priority based traffic switching is relevant only to Packet
          Switch Capable nodes. Nodes supporting this criteria will
          be able to interpret the EXP bits on the MPLS header to
          prioritize the traffic across the same LSP.

          Backup criteria refers to whether or not the node is capable
          of finding automatic protection path in the case the
          originally selected link fails. Such a local recovery is
          specific to the node and may not need to be notified to the
          upstream node.

      MPLS-Signaling protocol TLV (Tag ID = 3)
          The value can be 2 bytes long,  listing a combination of
          RSVP-TE, CR-LDP and LDP.

      Constraint-SPF algorithms-Support TLV (Tag ID = 4)
          List all the CSPF algorithms supported. Support for CSPF
          algorithms on a node is an indication that the node may be
          requested for all or partial circuit path selection during
          circuit setup time. Further, the CSPF algorithm support on
          an intermediate node can be beneficial when the node
          terminates one or more of the hierarchical LSP tunnels.

      Label Stack Depth TLV (Tag ID = 5)
          Applicable only for PSC-Type traffic. A default value of 1
          is assumed. This indicates the depth of label stack the
          node is capable of processing on an ingress interface.

   The following fields are used to describe each router link (i.e.,
   interface). Each router link is typed (see the below Type field).
   The Type field indicates the kind of link being described.

        A new link type "Positional-Ring Type" (value 5) is defined.
        This is essentially a connection to a TDM-Ring. TDM ring network
        is different from LAN/NBMA transit network in that, nodes on the
        TDM ring donot necessarily have a terminating path between
        themselves. Secondly, the order of links is important in
        determining the circuit path. Third, the protection switching
        and the number of fibers from a node going into a ring are
        determined by the ring characteristics. I.e., 2-fibre vs
        4-fibre ring and UPSR vs BLSR protected ring.

Srisuresh & Joseph                                             [Page 15]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

                 Type   Description
                 1      Point-to-point connection to another router
                 2      Connection to a transit network
                 3      Connection to a stub network
                 4      Virtual link
                 5      Positional-Ring Type.

   Link ID
        Identifies the object that this router link connects to.
        Value depends on the link's Type. For a positional-ring type,
        the Link ID shall be IP Network/Subnet number, just as with a
        broadcast transit network. The following table summarizes the
        updated Link ID values.

                       Type   Link ID
                       1      Neighboring router's Router ID
                       2      IP address of Designated Router
                       3      IP network/subnet number
                       4      Neighboring router's Router ID
                       5      IP network/subnet number

   Link Data
        This depends on the link's Type field. For type-5 links, this
        specifies the router interface's IP address.

   Link-TE options (TE capabilities of a link)

       |T|N|P|T|L|F|D|                                         |S|L|B|C|
       |E|T|K|D|S|S|B|                                         |R|U|W|O|
       | |E|T|M|C|C|S|                                         |L|G|A|L|
       |<---- Boolean TE flags ------->|<- TE flags pointing to TLVs ->|

       TE       - Indicates whether TE is permitted on the link. A link
                  can be denied for TE use by setting the flag to 0.

       NTE      - Indicates whether non-TE traffic is permitted on the
                  TE link. This flag is relevant only when the TE
                  flag is set.

       PKT      - Indicates whether or not the link is capable of
                  packet termination.

       TDM, LSC, FSC bits

Srisuresh & Joseph                                             [Page 16]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

                - Same as defined for router TE options.

       DBS      - Indicates whether or not Database synchronization
                  is permitted on this link.

       SRLG Bit - Shared Risk Link Group TLV follows.

       LUG  bit - Link usage cost metric TLV follows.

       BWA  bit - Data Link bandwidth TLV follows.

       COL bit  - Data link Color TLV follows.

   Link-TE TLVs

            This describes the list of Shared Risk Link Groups the link
            belongs to. Use 2 bytes to list each SRLG.

            This indicates the maximum bandwidth, available bandwidth,
            reserved bandwidth for later use etc. This TLV may also
            describe the Data link Layer protocols supported and the
            Data link MTU size.

           This indicates the link usage cost - Bandwidth unit, Unit
           usage cost, LSP setup cost, minimum and maximum durations
           permitted for setting up the TLV etc., including any time
           of day constraints.

           This is similar to the SRLG TLV, in that an autonomous
           system may choose to issue colors to link based on a
           certain criteria. This TLV can be used to specify the
           color assigned to the link within the scope of the AS.

7.2. Changes to Network LSAs

    Network-LSAs are the Type 2 LSAs. With the exception of the
    following, no additional changes will be required to this LSA
    for TE compatibility. The LSA format and flooding scope
    remains unchanged.

    A network-LSA is originated for each broadcast, NBMA and
    Positional-Ring type network in the area which supports two or
    more routers.  The TE option is also required to be set while
    propagating the TDM network LSA.

Srisuresh & Joseph                                             [Page 17]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

7.2.1. Positional-Ring type network LSA - New Network type for TDM-ring.
        - Ring ID: (Network Address/Mask)
        - No. of elements in the ring (a.k.a. ring neighbors)
        - Ring Bandwidth
        - Ring Protection (UPSR/BLSR)
        - ID of individual nodes (Interface IP address)
        - Ring type (2-Fibre vs. 4-Fibre, SONET vs. SDH)

    Network LSA will be required for SONET RING. Unlike the broadcast
    type, the sequence in which the NEs are placed on a RING-network
    is pertinent. The nodes in the ting must be described clock wise,
    assuming the GNE as the starting element.

7.3. TE-Summary LSAs

    TE-Summary-LSAs are the Type 0x83 and 0x84 LSAs. These LSAs are
    originated by area border routers. TE-Summary-network-LSA (0x83)
    describes the  reachability of TE networks in a non-backbone
    area, advertised by the Area Border Router. Type 0x84
    summary-LSA describes the reachability of Area Border Routers
    and AS border routers and their TE capabilities.

    One of the benefits of having multiple areas within an AS is
    that frequent TE advertisements within the area donot impact
    outside the area. Only the TE abstractions as befitting the
    external areas are advertised.

7.3.1. TE-Summary Network LSA (0x83)

    TE-summary network LSA may be used to advertise reachability of
    TE-networks accessible to areas external to the originating
    area. The scope of flooding is AS wide, with the exception of
    the originating area and the stub areas. For example, the
    TE-summary network LSA advertised by the border router of a
    non-backbone area is readvertised to all other areas, not just
    the backbone area. The area border router for each
    non-backbone area is responsible for advertising the
    reachability of backbone networks into the area.

    The flooding scope of TE-summary network LSA is unlike that
    of the summary network LSA (type 0x03), which simply uses this
    as an inter-area exchange of network accessibility and limits
    the flooding scope to just the backbone area.

        0                   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

Srisuresh & Joseph                                             [Page 18]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

       |            LS age             |     Options   |    0x83       |
       |                  Link State ID  (IP Network Number)           |
       |            Advertising Router (Area Border Router)            |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |                         Network Mask                          |
       |                         Area-ID                               |

7.3.2. TE-Summary router LSA (0x84)

    TE-summary router LSA may be used to advertise the availability of
    Area Border Routers (ABRs) and AS Border Routers (ASBRs) that are
    TE capable. The TE-summary router LSAs are originated by the Area
    Border Routers. The scope of flooding for the TE-summary router LSA
    is the entire AS, with the exception of the non-backbone areas the
    advertising ABRs belong to.

        0                   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
       |            LS age             |     Options   |      0x84     |
       |                     Link State ID                             |
       |                     Advertising Router (ABR)                  |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |    0      |E|B|      0        |       No. of Areas            |
       |                     Area-ID                                   |
       |                       ...                                     |
       |                   Router-TE flags                             |
       |                   Router-TE TLVs                              |

Srisuresh & Joseph                                             [Page 19]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

       |                     ....                                      |

   Link State ID
       The ID of the Area border router or the AS border router whose
       TE capability is being advertised.

   Advertising Router
       The ABR that advertises its TE capabilities (and the OSPF areas
       it belongs to) or the TE capabilities of an ASBR within one of
       the areas the ABR is a border router of.

   No. of Areas
       Specifies the number of OSPF areas the link state ID belongs to.

       Specifies the OSPF area(s) the link state ID belongs to. When
       the link state ID is same as the advertising router ID, this
       lists all the areas the ABR belongs to. In the case the
       link state ID is an ASBR, this simply lists the area the
       ASBR belongs to. The advertising router is assumed to be the
       ABR from the same area the ASBR is located in.

   Summary-router-TE flags

       Bit E - When set, the advertised Link-State ID is an AS boundary
               router (E is for external). The advertising router and
               the Link State ID belong to the same area.

       Bit B - When set, the advertised Link state ID is an Area
               border router (B is for Border)

   Router-TE flags,
   Router-TE TLVs  (TE capabilities of the link-state-ID router)

       TE Flags and TE TLVs are as applicable to the ABR/ASBR
       specified in the link state ID. The semantics is same as
       specified in the Router-TE LSA.

7.4. TE-AS-external LSAs (0x85)

    TE-AS-external-LSAs are the Type 0x85 LSAs. This is modeled after
    AS-external LSA format and flooding scope. These LSAs are originated
    by AS boundary routers with TE extensions (say, a BGP node which can
    communicate MPLS labels across to external ASes), and describe
    networks and pre-engineered TE links external to the AS. The

Srisuresh & Joseph                                             [Page 20]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

    flooding scope of this LSA is similar to that of an AS-external LSA.
    I.e., AS wide, with the exception of stub areas.

        0                   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
       |            LS age             |     Options   |      0x85     |
       |                        Link State ID                          |
       |                     Advertising Router                        |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |                         Network Mask                          |
       |                       Forwarding address                      |
       |                      External Route Tag                       |
       |  #  of Virtual TE links       |                 0             |
       |                      Link-TE flags                            |
       |                      Link-TE TLVs                             |
       |                              ...                              |
       |                      TE-Forwarding address                    |
       |                      External Route TE Tag                    |
       |                              ...                              |

    Network Mask
        The IP address mask for the advertised TE destination.  For
        example, this can be used to specify access to a specific
        TE-node or TE-link with an mask of 0xffffffff. This can also
        be used to specify access to an aggregated set of destinations
        using a different mask, ex: 0xff000000.

    Link-TE flags,
    Link-TE TLVs
        The TE attributes of this route. These fields are optional and
        are provided only when one or more pre-engineered circuits can

Srisuresh & Joseph                                             [Page 21]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

        be specified with the advertisement. Without these fields,
        the LSA will simply state TE reachability info.

    Forwarding address
        Data traffic for the advertised destination will be forwarded to
        this address.  If the Forwarding address is set to, data
        traffic will be forwarded instead to the LSA's originator (i.e.,
        the responsible AS boundary router).

    External Route Tag
        A 32-bit field attached to each external route.  This is not
        used by the OSPF protocol itself.  It may be used to communicate
        information between AS boundary routers; the precise nature of
        such information is outside the scope of this specification.

7.5. TE-Circuit-paths LSA (0x8C)

    TE-Circuit-paths LSA may be used to advertise the availability of
    pre-engineered TE circuit path(s) originating from any router in
    the network. The flooding scope may be Area wide or AS wide.

        0                   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
       |            LS age             |     Options   |      0x84     |
       |                       Link State ID                           |
       |                     Advertising Router                        |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |    0    |S|E|B|      0        |  #  of TE circuit paths       |
       |                          TE-Link ID                           |
       |                         TE-Link Data                          |
       |     Type      |        0      |    Link-TE flags              |
       |   Link-TE flags (contd.)      |  Zero or more Link-TE TLVs    |
       |                       TE-Link ID                              |
       |                      TE-Link Data                             |

Srisuresh & Joseph                                             [Page 22]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

       |                              ...                              |

   Link State ID
       The ID of the router to which the TE circuit path(s) is being

   TE-circuit-path(s) flags

       Bit S - When set, the flooding scope is set to be AS wide.
               Otherwise, the flooding scope is set to be area wide.

       Bit E - When set, the advertised Link-State ID is an AS boundary
               router (E is for external). The advertising router and
               the Link State ID belong to the same area.

       Bit B - When set, the advertised Link state ID is an Area border
               router (B is for Border)

   No. of Virtual TE Links
       This indicates the number of pre-engineered TE links between the
       advertising router and the router specified in the link state ID.

   TE-Link ID
       This is the ID by which to identify the virtual link on the
       advertising router. This can be any private interface index or
       handle that the advertising router uses to identify the
       pre-engineered TE virtual link to the ABR/ASBR.

   TE-Link Data
        This specifies the IP address of the physical interface
        on the advertising router.

7.6. TE-Link-Update LSA (0x8d)

   A significant difference between a non-TE OSPF network and a TE OSPF
   network is that the latter is subject to dynamic circuit pinning and
   is more likely to undergo state updates. Specifically, some links
   might undergo more changes and more frequently than others.
   Advertising the entire TE-router LSA in response to a change in any
   single link could be repetitive. Flooding the network with TE-router
   LSAs at the aggregated speed of all the dynamic changes is simply
   not desirable. Hence, the new TE-link-update LSA, that advertises
   link specific updates alone.

   The TE-link-Update LSA will be advertised as frequently as the link
   state is changed. The TE-link sequence is largely the advertisement
   of a sub-portion of router LSA. The sequence number on this will be

Srisuresh & Joseph                                             [Page 23]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   incremented with the TE-router LSA's sequence as the basis. When an
   updated TE-router LSA is advertised within 30 minutes of the
   previous advertisement, the updated TE-router LSA will assume a
   sequence no. that is larger than the most frequently updated of
   its links.

   Below is the format of the TE-link update LSA.

        0                   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
       |            LS age             |     Options   |     0x8d      |
       |                        Link State ID (same as Link ID)        |
       |                     Advertising Router                        |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |                         Link Data                             |
       |     Type      |        0      |    Link-TE options            |
       |     Link-TE options           | Zero or more Link-TE TLVs     |
       |     # TOS     |                            metric             |
       |                              ...                              |
       |      TOS      |        0      |          TOS  metric          |

   Link State ID
       This would be exactly the same as would have been specified as
       as Link ID for a link within the router-LSA.

   Link Data
       This specifies the router ID the link belongs to. In majority of
       cases, this would be same as the advertising router.

   The tuple of (LS Type, LSA ID, Advertising router) uniquely identify
   the LSA and replace LSAs of the same tuple with an older sequence
   number. However, there is an exception to this rule in the context
   of TE-link-update LSA. TE-Link update LSA will initially assume the
   sequence number of the TE-router LSA it belongs to. Further,

Srisuresh & Joseph                                             [Page 24]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   when a new TE-router LSA update with a larger sequence number is
   advertised, the newer sequence number is assumed by al the link

7.7. TE-Router-Proxy LSA (0x8e)

   This is a variation to the TE-router LSA in that the TE-router LSA
   is not advertised by the network element, but rather by a trusted
   TE-router Proxy. This is typically the scenario in a non-packet
   TE network, where some of the nodes donot have OSPF functionality
   and count on a helper node to do the advertisement for them. One
   such example would be the SONET/SDH ADM nodes in a TDM ring. The
   nodes may principally depend upon the GNE (Gateway Network Element)
   to do the advertisement for them. TE-router-Proxy LSA shall not be
   used to advertise Area Border Routers and/or AS border Routers.

        0                   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
       |            LS age             |     Options   |     0x8e      |
       |      Link State ID  (Router ID of the TE Network Element)     |
       |                     Advertising Router                        |
       |                     LS sequence number                        |
       |         LS checksum           |             length            |
       |                 0             |       Router-TE flags         |
       |  Router-TE flags (contd.)     |       Router-TE TLVs          |
       |                     ....                                      |
       |                     ....      |      # of TE links            |
       |                          Link ID                              |
       |                         Link Data                             |
       |     Type      |        0      |    Link-TE options            |
       |   Link-TE flags               |  Zero or more Link-TE TLVs    |
       |                          Link ID                              |

Srisuresh & Joseph                                             [Page 25]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

       |                         Link Data                             |
       |                              ...                              |

8. Abstract topology representation with TE support

   Below, we assume a TE network that is composed of three OSPF areas,
   namely Area-1, Area-2 and Area-3, attached together through the
   backbone area. The following figure is an inter-area topology
   abstraction from the perspective of routers in Area-1. The
   abstraction is similar, but not the same, as that of the non-TE
   abstraction. As such, the authors claim the model is easy to
   understand and emulate. The abstraction illustrates reachability
   of TE networks and nodes in areas external to the local area and
   ASes external to the local AS. The abstraction also illustrates
   pre-engineered TE links that may be advertised by ABRs and ASBRs.

   Area-1 an has a single border router, ABR-A1 and no ASBRs. Area-2
   has an Area border router ABR-A2 and an AS border router ASBR-S1.
   Area-3 has two Area border routers ABR-A2 and ABR-A3; and an AS
   border router ASBR-S2. There may be any number of Pre-engineered
   TE links amongst ABRs and ASBRs. The following example assumes a
   single TE-link between ABR-A1 and ABR-A2; between ABR-A1 and
   ABR-A3; between ABR-A2 to ASBR-S1; and between ABR-A3 to ASBR-S2.
   All Area border routers and AS border routers are assumed to
   be represented by their TE capabilities.

Srisuresh & Joseph                                             [Page 26]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

                           |Area-1 |
    +-------------+            |
    |Reachable TE |        +------+
    |networks in  |--------|ABR-A1|
    |backbone area|        +------+
    +-------------+         | | |
              +-------------+ | +-------------------+
              |                 |                   |
    +-----------------+       |             +-----------------+
    |Pre-engineered TE|    +----------+     |Pre-engineered TE|
    |circuit path(s)  |    | Backbone |     |circuit path(s)  |
    |to ABR-A2        |    | Area     |     |to ABR-A3        |
    +-----------------+    +----------+     +-----------------+
              |               |   |                 |
              +----------+    |   |                 |
                         |    |   +--------------+  |
    +-----------+        |    |                  |  |     +-----------+
    |Reachable  |   +------------+              +------+  |Reachable  |
    |TE networks|---|    ABR-A2  |              |ABR-A3|--|TE networks|
    |in Area A2 |   +------------+              +------+  |in Area A3 |
    +-----------+   /   | | |                     | |     +-----------+
                  /     | | +-------------------+ | +----------+
                /       | +-------------+       | |            |
    +-----------+ +--------------+      |       | |    +--------------+
    |Reachable  | |Pre-engineered|      |       | |    |Pre-engineered|
    |TE networks| |TE Ckt path(s)|  +------+  +------+ |TE Ckt path(s)|
    |in Area A3 | |to ASBR-S1    |  |Area-2|  |Area-3| |to ASBR-S2    |
    +-----------+ +--------------+  +------+  +------+ +--------------+
                             |      /              |        /
    +-------------+          |     /               |       /
    |AS external  |      +---------+          +-------------+
    |TE-network   |------| ASBR-S1 |          |  ASBR-S2    |
    |reachability |      +---------+          +-------------+
    |from ASBR-S1 |         |                   |        |
    +-------------+         |                   |        |
            +-----------------+   +-------------+   +-----------------+
            |Pre-engineered TE|   |AS External  |   |Pre-engineered TE|
            |circuit path(s)  |   |TE-Network   |   |circuit path(s)  |
            |reachable from   |   |reachability |   |reachable from   |
            |ASBR-S1          |   |from ASBR-S2 |   |ASBR-S2          |
            +-----------------+   +-------------+   +-----------------+

   Figure 8: Inter-Area Abstraction as viewed by Area-1 TE-routers

Srisuresh & Joseph                                             [Page 27]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

9. Changes to Data structures in OSPF-TE routers

9.1. Changes to Router data structure

   The router with TE extensions must be able to include all the
   TE capabilities (as specified in section 7.1) in the router data
   structure. Further, routers providing proxy service to other TE
   routers must also track the router and associated interface data
   structures for all the TE client nodes for which the proxy
   service is being provided. Presumably, the interaction between
   the Proxy server and the proxy clients is out-of-band.

9.2. Two set of Neighbors

   Two sets of neighbor data structures will need to be maintained.
   TE-neighbors set is used to advertise TE LSAs. Only the
   TE-routers will be members of the TE-neighbor set.
   Normal neighbors set will be used to advertise native LSAs. All
   neighboring nodes supporting non-TE links canbe part of this
   set. As for flooding optimizations based on neighbors set,
   readers may refer [OSPF-FL1].

9.3. Changes to Interface data structure

   The following new fields are introduced to the interface data
   structure. These changes are in addition to the changes specified
   in [OSPF-FL1].

       If the value of the flag is TRUE, the interface is permissible
       to be advertised as a TE-enabled interface.

       If the value of the flag is TRUE, the interface permits non-TE
       traffic on the interface. Specifically, this is applicable to
       packet networks, where data links may permit both TE and non-TE
       packets. For FSC and LSC TE networks, this flag will be set to
       FALSE. For Packet networks that donot permit non-TE traffic on
       TE links alos, this flag is set to TRUE.

       If the value of the flag is TRUE, the interface terminates
       Packet data and hence may be used for IP and OSPF data exchange.

       If the value of the flag is TRUE, the interface may be used to
       synchronize the LSDB across all adjacent neighbors. This is
       TRUE by default to all PktTerminated interfaces that are

Srisuresh & Joseph                                             [Page 28]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

       enabled for OSPF. However, it is possible to set this to FALSE
       for some of the interfaces.

       Each interface may potentially have a maximum of 16 TLVS that
       describe the link characteristics.

   The following existing fields in Interface data structure will take
   on additional values to support TE extensions.

       The OSPF interface type can also be of type "Positional-RING".
       The Positional-ring type is different from other types (such
       as broadcast and NBMA) in that the exact location of the nodes
       on the ring is relevant, even as they are all on the same
       ring. SONET ADM ring is a good example of this. Complete ring
       positional-ring description may be provided by the GNE on a
       ring as a TE-network LSA for the ring.

   List of Neighbors
       The list may be statically defined for an interface, without
       requiring the use of Hello protocol.

10. Comparison between Opaque-LSAs & the new TE-LSAs

   The following subsections attempt to identify the various issues
   such as flooding scope and scalability that are fundamentally
   lacking in the Opage-LSA based approach. Section 10.2 goes on to
   describe a transition strategy to eventually transition completely
   to the new TE-LSA scheme.

   Once the OSPF-TE is completely transitioned to the scheme
   described in this document, the packet and non-packet networks
   can be combined and issued addresses across the unified network.
   As such, the traffic engineering can be based on the overlayed or
   the peer model espoused in [GMPLS-TE].

10.1. TE flooding load on a non-TE network

   In a non-TE network, when a link is flapping, that can cause
   considerable hardship on all routers in the area. The hardship is
   not so much because of the LSAs that are generated, but because
   that causes the OSPF routing table to be recalculated.

   A TE network can also have a large number of LSA updates due to
   the many state changes the TE links undergo dynamically. These
   LSA updates are neither infrequent nor undesirable as with link

Srisuresh & Joseph                                             [Page 29]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001


   Now, consider the case where Opaque LSAs are used for TE
   extensions. The flooding topology for opaque LSAs makes no
   distinction between TE nodes and non-TE nodes. In a network
   where the TE and non-TE nodes coexist, a non-TE router would be
   bombarded with Opaque LSAs.

   These Opaque LSAs carry TE metric state changes, which the non-TE
   router does not care about. If the router simply dropped the opaque
   LSAs and didnt recompute the dijkstra, that might be OK. But, it
   may be that some routers will recompute routes (because they process
   some of the Opaque LSAs that say that a particular link is no longer
   available for non-TE use). In the latter case, the routers might
   choose to simply recompute for all new Opaque LSA advertisements.
   Clearly, that would be a considerable computational demand and
   cause for instability on non-TE routers, triggered by the frequent
   opaque LSA advertisements.

   Secondly, If the TE and non-TE topologies are not separated (as is
   the case with Opaque-LSAs), the non-TE router could be utilizing the
   TE link as its least cost link, thereby stressing the TE link and
   effectively rendering the TE link  ineffective for TE purposes.
   Separating the two topologies (as advocated by this document with
   new TE LSAs and TE option flag) ensure that the SLA objectives on
   TE links are properly met.

   Thirdly, the wider the flooding scope, the larger the number of
   retransmissions and acknowledgements. The same information or
   sometimes unneeded information may reach a router through multiple
   links. Even if the router didnt forward the information past the
   time, it would still have to send acknowledgements across all the
   multiple links on which the LSAs tried to converge. By moving the
   concept of flooding from "per interface" to "per neighbor", we
   minimize the flooding, without compromising on the untimate goal
   of LSDB convergence for TE and non-TE networks.

   Lastly, separating TE and non-TE topologies is beneficial in
   inter-area communication. When the topologies are separate, the
   area border routers can advertise different summary LSAs for TE and
   non-TE routers. Opaque LSAs are not adequate to establish
   TE peering relationship with the neighbors.

   For example, a non-TE Area Border router (ABR) could simply announce
   the non-TE-network summary LSAs (LSA type 3) for non-TE networks
   outside the area. A TE ABR, on the other hand, could advertise
   just the TE-network summary LSAs (0x83). Clearly, the advertised
   data is different. The boundary of TE-network summary LSA flooding

Srisuresh & Joseph                                             [Page 30]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   is also different. The flooding boundary for TE-summary LSAs would
   be (AS - OriginatingArea - StubAreas - NSSAs). Clearly, the
   Opaque-LSA flooding boundary will not permit this type of flooding
   granularity. Without an AS-wide flooding (with the exception of stub
   areas), it is impossible to know which outside-are networks are
   TE-configurable and which are not.

   In summary, lack of flexible flooding topology can be an operational
   and functional nightmare. Folks will be forced to an unscalable,
   single-area topology to get around the shortcoming of the opaque

10.2. Scaling concerns What is lacking in Opaque-LSA-based TE scheme?

   The Opaque LSA based mechanism has the following fundamental scaling
   problems. These cannot be fixed by mere extensions to the same
   approach. We suggest below a transition strategy to migrate to the
   scheme proposed in this document.

   1. The flooding boundaries of Opaque LSAs make the OSPF-TE suitable
      at best to single-area topologies. Extending TE beyond one area
      can cause a lot of flooding problems. e.g.: Opaque LSAs cannot
      support the flooding scope of TE-summary-networks.
      Opaque LSAs (AS-wide scope) will be unable to restrict flooding
      in its own originating area.

   2. The Opaque LSA is also restricted in the way it can express
      different types of data. Everything should be expressible in
      in the form of a TLV. Summary-TE-networks-from each-Area,
      TE-ABR routers, TE-ASBR routers, TE-AS-External-networks,
      TE-Router-Capabilities, TE-link-updates, Pre-engineered-TE-Links
      - All of these data have to be engineered to be expressible in a
      TLV form with one or more sub-TLVs. TLVs should not be a panacea
      for all kinds of TE data. TLVs are generally more difficult to
      process and debug than fixed format messages.

      Opaque LSAs demand more processing to assimilate into topology
      abstraction. A single Opaque LSA type is bent in many
      ways (using a variety of TLVs) to update the native OSPF topology
      abstraction nodes.

   One way to transition from the current Opaque-LSA-based TE scheme to
   the new-TE scheme could be as follows.

   1. Use the existing Opaque-LSA-based-TE scheme for single area
      topologies. You will still need to find a way that a non-TE
      router doesnt cannibalize a TE-link for SPF forwarding.

Srisuresh & Joseph                                             [Page 31]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

   2. Fold in the TE option flag to construct the TE and non-TE
      topologies in an area, even if the topologies cannot be used
      for flooding within the area.

   3. Do away with Opaque LSAs for inter-area communication. Make use
      of the TE-topology within area to summarize the TE networks in
      the area and advertise the same to all TE-routers in the backbone.
      The TE-ABRs on the backbone area will in-turn advertise these
      summaries again within their connected areas. Use new LS types
      for summary LSAs, AS-external-LSAs and so forth, as specified
      in this document.

   4. Replace the use of Opaque LSAs with the TE LSAs within the area
      as well.

10.3. Link State Database.

   With the new TE-LSA scheme, a TE node will have two types of
   Link state databases. The normal LSDB describes the control
   (non-TE) topology. Shortest-Path-First algorithm will be used to
   forward IP packets along this network. OSPF neighbors data
   structure will be used for flooding along the control topology.

   The TE node will have a separate TE-LSDB that describes the TE
   topology, constituted only of TE nodes and TE links. A variety of
   CSPF algorithms may be used to dynamically setup TE circuit paths
   along this TE network. TE-neighbors data structure is used for
   flooding TE LSAs alongs the TE-only topology. Having a clear
   distinction between the two LSDBs (and hence topologies) makes
   this approach more desirable to service providers desiring to
   offer strictly enforceable SLAs (Service Level Agreements)
   along their TE topology.

   Whereas, in the Opaque-LSA-based TE scheme, the TE-LSDB built
   using opaque LSAs will be required to refer the normal LSDB to
   build the TE topology. Even with that, there is way to know the
   TE capabilities of the routers. The Opaque-LSA approach does
   not deal with TE capabilities for a router. Opaque LSAs
   are flooded to all nodes. Some nodes that happen to support
   the TE extensions will have a hit and accept the opaque LSAs.
   Others that donot support will have a miss and simply drop the
   received Opaque LSAs. This type of hit-and-miss approach is
   not only disruptive, but also blind to the SLA requirements
   on TE links.

Srisuresh & Joseph                                             [Page 32]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

10.4. Real-world scenarios better served by the new-TE-LSAs scheme.

   Many real-world scenarios are better served by the new-TE-LSAs
   scheme. Here are a few examples.

   1. Multi-area network.

   2. Single-Area networks - The TE links are not cannibalized by the
      non-TE routers for SPF forwarding.

   3. Credible SLA enforcement in a (TE + non-TE) packet network.
      Ability to restrict flooding to some links (say, non-TE links)
      ensures the service provider is able to devote the entire
      bandwidth of a TE-link for TE circuit purposes. This makes SLA
      enforcement credible.

   4. For a non-Packet TE network, the Opaque-LSA-based-TE scheme is
      not adequate to represent
      (a) "Positional-Ring" type network LSA and
      (b) Router Proxying - allowing a router to advertise on behalf
          of other nodes (that are not Packet/OSPF capable).

11. IANA Considerations

11.1. All-TE-compliant-SPF routers Multicast address allocation

11.2. New TE-LSA Types

11.3. New TLVs (Router-TE and Link-TE TLVs)

11.3.1. TE-selection-Criteria TLV (Tag ID = 1)
          - Bandwidth based LSPs (1)
          - Priority based LSPs  (2)
          - Backup LSP           (3)
          - Link cost            (4)

11.3.2. MPLS-Signaling protocol TLV (Tag ID = 3)
          - RSVP-TE signaling
          - LDP signaling
          - CR-LDP signaling

11.3.3. Constraint-SPF algorithms-Support TLV (Tag ID = 4)
          - CSPF Algorithm Codes.

11.3.4. SRLG-TLV (Tag ID = 0x81)
          - SRLG group IDs

Srisuresh & Joseph                                             [Page 33]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

11.3.5. BW-TLV (Tag ID = 0x82)

11.3.6  CO-TLV (Tag ID = ox83)

12. Security Considerations

    This memo does not create any new security issues for the OSPF
    protocol. Security considerations for the base OSPF protocol are
    covered in [OSPF-v2]. As a general rule, a TE network is likely
    to generate significantly more control traffic than a native
    OSPF network. The excess traffic is almost directly proportional
    to the rate at which TE circuits are setup and torn down within
    an autonomous system. It is important to ensure that TE database
    sychronizations happen quickly when compared to the aggregate
    circuit setup an tear-down rates.

Srisuresh & Joseph                                             [Page 34]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001


   [IETF-STD] Bradner, S., " The Internet Standards Process --
              Revision 3", RFC 1602, IETF, October 1996.

   [RFC 1700] J. Reynolds and J. Postel, "Assigned Numbers",
              RFC 1700

   [MPLS-TE]  Awduche, D., et al, "Requirements for Traffic
              Engineering Over MPLS," RFC 2702, September 1999.

   [GMPLS-TE] P.A. Smith et. al, "Generalized MPLS - Signaling
              Functional Description",
              draft-ietf-mpls-generalized-signaling-03.txt, work
              in progress.

   [RSVP-TE]  Awduche, D.O., L. Berger, Der-Hwa Gan, T. Li,
              V. Srinivasan and G. Swallow, "RSVP-TE: Extensions
              to RSVP for LSP Tunnels", Work in progress,

   [CR-LDP]   Jamoussi, B. et. al, "Constraint-Based LSP Setup
              using LDP", draft-ietf-mpls-cr-ldp-05.txt,
              Work in Progress.

   [OSPF-v2]  Moy, J., "OSPF Version 2", RFC 2328, April 1998.

   [MOSPF]    Moy, J., "Multicast Extensions to OSPF", RFC 1584,
              March 1994.

   [NSSA]     Coltun, R., V. Fuller and P. Murphy, "The OSPF NSSA
              Option", draft-ietf-ospf-nssa-update-10.txt, Work in

   [OPAQUE]   Coltun, R., "The OSPF Opaque LSA Option," RFC 2370,
              July 1998.

   [OSPF-FL1] Zinin, A. and M. Shand, "Flooding Optimizations in
              link-state routing protocols", work in progress,

   [OSPF-FL2] Moy, J., "Flooding over a subset topology",
              <draft-ietf-ospf-subset-flood-00.txt>, work in progress.

   [OPQLSA-TE] Katz, D., D. Yeung and K. Kompella, "Traffic
              Engineering Extensions to OSPF", work in progress,

Srisuresh & Joseph                                             [Page 35]

Internet-Draft       TE LSAs for OSPF-TE extensions            June 2001

Authors' Addresses

   Pyda Srisuresh
   Jasmine Networks
   3061 Zanker Road, Suite B
   San Jose, CA 95134
   EMail: srisuresh@yahoo.com

   Paul Joseph
   Jasmine Networks
   3061 Zanker Road, Suite B
   San Jose, CA 95134
   EMail: pjoseph@jasminenetworks.com

Srisuresh & Joseph                                             [Page 36]

Html markup produced by rfcmarkup 1.108, available from http://tools.ietf.org/tools/rfcmarkup/