RFC 8243






Internet Engineering Task Force (IETF)                        R. Perlman
Request for Comments: 8243                                           EMC
Category: Informational                                  D. Eastlake 3rd
ISSN: 2070-1721                                                 M. Zhang
                                                                  Huawei
                                                             A. Ghanwani
                                                                    Dell
                                                                 H. Zhai
                                                                     JIT
                                                          September 2017


                      Alternatives for Multilevel
          Transparent Interconnection of Lots of Links (TRILL)

Abstract



   Although TRILL is based on IS-IS, which supports multilevel unicast
   routing, extending TRILL to multiple levels has challenges that are
   not addressed by the already-existing capabilities of IS-IS.  One
   issue is with the handling of multi-destination packet distribution
   trees.  Other issues are with TRILL switch nicknames.  How are such
   nicknames allocated across a multilevel TRILL network?  Do nicknames
   need to be unique across an entire multilevel TRILL network?  Or can
   they merely be unique within each multilevel area?

   This informational document enumerates and examines alternatives
   based on a number of factors including backward compatibility,
   simplicity, and scalability; it makes recommendations in some cases.

Status of This Memo



   This document is not an Internet Standards Track specification; it is
   published for informational purposes.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Not all documents
   approved by the IESG are a candidate for any level of Internet
   Standard; see Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc8243.






Perlman, et al.               Informational                     [Page 1]

RFC 8243              Multilevel TRILL Alternatives       September 2017


Copyright Notice



   Copyright (c) 2017 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (https://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.





































Perlman, et al.               Informational                     [Page 2]

RFC 8243              Multilevel TRILL Alternatives       September 2017


Table of Contents



   1. Introduction ....................................................4
      1.1. The Motivation for Multilevel ..............................4
      1.2. Improvements Due to Multilevel .............................5
           1.2.1. The Routing Computation Load ........................5
           1.2.2. LSDB Volatility Creating Too Much Control Traffic ...5
           1.2.3. LSDB Volatility Causing Too Much Time Unconverged ...6
           1.2.4. The Size of the LSDB ................................6
           1.2.5. Nickname Limit ......................................6
           1.2.6. Multi-Destination Traffic ...........................7
      1.3. Unique and Aggregated Nicknames ............................7
      1.4. More on Areas ..............................................8
      1.5. Terminology and Abbreviations ..............................9
   2. Multilevel TRILL Issues ........................................10
      2.1. Non-Zero Area Addresses ...................................11
      2.2. Aggregated versus Unique Nicknames ........................12
           2.2.1. More Details on Unique Nicknames ...................12
           2.2.2. More Details on Aggregated Nicknames ...............13
      2.3. Building Multi-Area Trees .................................18
      2.4. The RPF Check for Trees ...................................18
      2.5. Area Nickname Acquisition .................................19
      2.6. Link State Representation of Areas ........................19
   3. Area Partition .................................................20
   4. Multi-Destination Scope ........................................21
      4.1. Unicast to Multi-Destination Conversions ..................21
           4.1.1. New Tree Encoding ..................................22
      4.2. Selective Broadcast Domain Reduction ......................22
   5. Coexistence with Old TRILL Switches ............................23
   6. Multi-Access Links with End Stations ...........................24
   7. Summary ........................................................25
   8. Security Considerations ........................................26
   9. IANA Considerations ............................................26
   10. References ....................................................26
      10.1. Normative References .....................................26
      10.2. Informative References ...................................27
   Acknowledgements ..................................................28
   Authors' Addresses ................................................29













Perlman, et al.               Informational                     [Page 3]

RFC 8243              Multilevel TRILL Alternatives       September 2017


1.  Introduction



   The IETF Transparent Interconnection of Lot of Links (TRILL) protocol
   [RFC6325] [RFC7177] [RFC7780] provides optimal pairwise data routing
   without configuration, safe forwarding even during periods of
   temporary loops, and support for multipathing of both unicast and
   multicast traffic in networks with arbitrary topology and link
   technology, including multi-access links.  TRILL accomplishes this by
   using Intermediate System to Intermediate System [IS-IS] [RFC7176])
   link state routing in conjunction with a header that includes a hop
   count.  The design supports Data Labels (VLANs and Fine-Grained
   Labels (FGLs) [RFC7172]) and optimization of the distribution of
   multi-destination data based on Data Label and multicast group.
   Devices that implement TRILL are called TRILL Switches or RBridges.

   Familiarity with [IS-IS], [RFC6325], and [RFC7780] is assumed in this
   document.

1.1.  The Motivation for Multilevel



   The primary motivation for multilevel TRILL is to improve
   scalability.  The following issues might limit the scalability of a
   TRILL-based network:

   1.  The routing computation load

   2.  The volatility of the link state database (LSDB) creating too
       much control traffic



   3.  The volatility of the LSDB causing the TRILL network to be in an
       unconverged state too much of the time



   4.  The size of the LSDB



   5.  The limit of the number of TRILL switches, due to the 16-bit
       nickname space (for further information on why this might be a
       problem, see Section 1.2.5)

   6.  The traffic due to upper-layer protocols use of broadcast and
       multicast



   7.  The size of the end-node learning table (the table that remembers
       (egress TRILL switch, label / Media Access Control (MAC)) pairs)



   As discussed below, extending TRILL IS-IS to be multilevel
   (hierarchical) can help with all of these issues except issue 7.





Perlman, et al.               Informational                     [Page 4]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   IS-IS was designed to be multilevel [IS-IS].  A network can be
   partitioned into "areas".  Routing within an area is known as "Level
   1 routing".  Routing between areas is known as "Level 2 routing".
   The Level 2 IS-IS network consists of Level 2 routers and links
   between the Level 2 routers.  Level 2 routers may participate in one
   or more Level 1 areas, in addition to their role as Level 2 routers.

   Each area is connected to Level 2 through one or more "border
   routers", which participate both as a router inside the area, and as
   a router inside the Level 2 area.  Care must be taken that it is
   clear, when transitioning multi-destination packets between a Level 2
   and a Level 1 area in either direction, that exactly one border TRILL
   switch will transition a particular data packet between the levels;
   otherwise, duplication or loss of traffic can occur.

1.2.  Improvements Due to Multilevel



   Partitioning the network into areas directly solves the first four
   scalability issues listed above, as described in Sections 1.2.1
   through 1.2.4.  Multilevel also contributes to solving issues 5 and
   6, as discussed in Sections 1.2.5 and 1.2.6, respectively.

   In the subsections below, N indicates the number of TRILL switches in
   a TRILL campus.  For simplicity, it is assumed that each TRILL switch
   has k links to other TRILL switches.  An "optimized" multilevel
   campus is assumed to have Level 1 areas containing sqrt(N) switches.

1.2.1.  The Routing Computation Load



   The Dijkstra algorithm uses computational effort on the order of the
   number of links in a network (N*k) times the log of the number of
   nodes to calculate least cost routes at a router (Section 12.3.3 of
   [InterCon]).  Thus, in a single-level TRILL campus, it is on the
   order of N*k*log(N).  In an optimized multilevel campus, it is on the
   order of sqrt(N)*k*log(N).  So, for example, assuming N is 3,000, the
   level of computational effort would be reduced by about a factor of
   50.

1.2.2.  LSDB Volatility Creating Too Much Control Traffic



   The rate of LSDB changes is assumed to be approximately proportional
   to the number of routers and links in the TRILL campus or N*(1+k) for
   a single-level campus.  With an optimized multilevel campus, each
   area would have about sqrt(N) routers and proportionately fewer links
   reducing the rate of LSDB changes by about a factor of sqrt(N).






Perlman, et al.               Informational                     [Page 5]

RFC 8243              Multilevel TRILL Alternatives       September 2017


1.2.3.  LSDB Volatility Causing Too Much Time Unconverged



   With the simplifying assumption that routing converges after each
   topology change before the next such change, the fraction of time
   that routing is unconverged is proportional to the product of the
   rate of change occurrence and the convergence time.  The rate of
   topology changes per some arbitrary unit of time will be roughly
   proportional to the number of router and links (Section 1.2.2).  The
   convergence time is approximately proportional to the computation
   involved at each router (Section 1.2.1).  Thus, based on these
   simplifying assumptions, the time spent unconverged in a single-level
   network is proportional to (N*(1+k))*(N*k*log(N)) while that time for
   an optimized multilevel network would be proportional to
   (sqrt(N)*(1+k))*(sqrt(N)*k*log(N)).  Thus, in changing to multilevel,
   the time spent unconverged, using these simplifying assumptions, is
   improved by about a factor of N.

1.2.4.  The Size of the LSDB



   The size of the LSDB, which consists primarily of information about
   routers (TRILL switches) and links, is also approximately
   proportional to the number of routers and links.  So, as with item 2
   in Section 1.2.2, it should improve by about a factor of sqrt(N) in
   going from single level to multilevel.

1.2.5.  Nickname Limit



   For many TRILL protocol purposes, RBridges are designated by 16-bit
   nicknames.  While some values are reserved, this appears to provide
   enough nicknames to designated over 65,000 RBridges.  However, this
   number is effectively reduced by the following two factors:

   -  Nicknames are consumed when pseudo-nicknames are used for the
      active-active connection of end stations.  Using the techniques in
      [RFC7781], for example, could double the nickname consumption if
      there are extensive active-active edge groups connected to
      different sets of edge TRILL switch ports.

   -  There might be problems in multilevel campus-wide contention for
      single-nickname allocation of nicknames were allocated
      individually from a single pool for the entire campus.  Thus, it
      seems likely that a hierarchical method would be chosen where
      blocks of nicknames are allocated at Level 2 to Level 1 areas and
      contention for a nickname by an RBridge in such a Level 1 area
      would be only within that area.  Such hierarchical allocation
      leads to further effective loss of nicknames similar to the
      situation with IP addresses discussed in [RFC3194].




Perlman, et al.               Informational                     [Page 6]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   Even without the above effective reductions in nickname space, a very
   large multilevel TRILL campus, say one with 200 areas each containing
   500 TRILL switches, could require 100,000 or more nicknames if all
   nicknames in the campus must be unique, which is clearly impossible
   with 16-bit nicknames.

   This scaling limit, namely, the 16-bit nickname space, will only be
   addressed with the aggregated-nickname approach.  Since the
   aggregated-nickname approach requires some complexity in the border
   TRILL switches (for rewriting the nicknames in the TRILL header), the
   suggested design in this document allows a campus with a mixture of
   unique-nickname areas, and aggregated-nickname areas.  Thus, a TRILL
   network could start using multilevel with the simpler unique nickname
   method and later add aggregated-nickname areas as a later stage of
   network growth.

   With this design, nicknames must be unique across all Level 2 and
   unique-nickname area TRILL switches taken together; whereas nicknames
   inside an aggregated-nickname area are visible only inside that area.
   Nicknames inside an aggregated-nickname area must still not conflict
   with nicknames visible in Level 2 (which includes all nicknames
   inside unique nickname areas), but the nicknames inside an
   aggregated-nickname area may be the same as nicknames used within one
   or more other aggregated-nickname areas.

   With the design suggested in this document, TRILL switches within an
   area need not be aware of whether they are in an aggregated-nickname
   area or a unique nickname area.  The border TRILL switches in area A1
   will indicate, in their LSP inside area A1, which nicknames (or
   nickname ranges) are or are not available to be chosen as nicknames
   by area A1 TRILL switches.

1.2.6.  Multi-Destination Traffic



   In many cases, scaling limits due to protocol use of broadcast and
   multicast can be addressed in a multilevel campus by introducing
   locally scoped multi-destination delivery, limited to an area or a
   single link.  See further discussion of this issue in Section 4.2.

1.3.  Unique and Aggregated Nicknames



   We describe two alternatives for hierarchical or multilevel TRILL.
   One we call the "unique-nickname" alternative.  The other we call the
   "aggregated-nickname" alternative.  In the aggregated-nickname
   alternative, border TRILL switches replace either the ingress or
   egress nickname field in the TRILL header of unicast packets with an
   aggregated nickname representing an entire area.




Perlman, et al.               Informational                     [Page 7]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   The unique-nickname alternative has the advantage that border TRILL
   switches are simpler and do not need to do TRILL Header nickname
   modification.  It also simplifies testing and maintenance operations
   that originate in one area and terminate in a different area.

   The aggregated-nickname alternative has the following advantages:

      -  it solves scaling issue 5 above, the 16-bit nickname limit, in
         a simple way,

      -  it lessens the amount of inter-area routing information that
         must be passed in IS-IS, and

      -  it logically reduces the RPF (Reverse Path Forwarding) Check
         information (since only the area nickname needs to appear,
         rather than all the ingress TRILL switches in that area).

   In both cases, it is possible and advantageous to compute multi-
   destination data packet distribution trees such that the portion
   computed within a given area is rooted within that area.

   For further discussion of the unique and aggregated-nickname
   alternatives, see Section 2.2.

1.4.  More on Areas



   Each area is configured with an "area address", which is advertised
   in IS-IS messages, so as to avoid accidentally interconnecting areas.
   For TRILL, the only purpose of the area address would be to avoid
   accidentally interconnecting areas although the area address had
   other purposes in CLNP (ConnectionLess Network Protocol), IS-IS was
   originally designed for CLNP/DECnet.

   Currently, the TRILL specification says that the area address must be
   zero.  If we change the specification so that the area address value
   of zero is just a default, then most IS-IS multilevel machinery works
   as originally designed.  However, there are TRILL-specific issues,
   which we address in Section 2.1.













Perlman, et al.               Informational                     [Page 8]

RFC 8243              Multilevel TRILL Alternatives       September 2017


1.5.  Terminology and Abbreviations



   This document generally uses the abbreviations defined in [RFC6325]
   plus the additional abbreviation DBRB.  However, for ease of
   reference, most abbreviations used are listed here:

   CLNP:          ConnectionLess Network Protocol

   DECnet:        a proprietary routing protocol that was used by
                  Digital Equipment Corporation.  "DECnet Phase 5" was
                  the origin of IS-IS.

   Data Label:    VLAN or Fine-Grained Label [RFC7172]

   DBRB:          Designated Border RBridge

   ESADI:         End-Station Address Distribution Information

   IS-IS:         Intermediate System to Intermediate System [IS-IS]

   LSDB:          Link State DataBase

   LSP:           Link State PDU

   PDU:           Protocol Data Unit

   RBridge:       Routing Bridge, an alternative name for a TRILL switch

   RPF:           Reverse Path Forwarding

   TLV:           Type-Length-Value

   TRILL:         Transparent Interconnection of Lots of Links or
                  Tunneled Routing in the Link Layer [RFC6325] [RFC7780]

   TRILL switch:  a device that implements the TRILL protocol [RFC6325]
                  [RFC7780], sometimes called an RBridge

   VLAN:          Virtual Local Area Network












Perlman, et al.               Informational                     [Page 9]

RFC 8243              Multilevel TRILL Alternatives       September 2017


2.  Multilevel TRILL Issues

   The TRILL-specific issues introduced by multilevel include the
   following:

   a.  Configuration of non-zero area addresses, encoding them in IS-IS
       PDUs, and possibly interworking with old TRILL switches that do
       not understand non-zero area addresses.

       See Section 2.1.

   b.  Nickname management.

       See Sections 2.5 and 2.2.

   c.  Advertisement of pruning information (Data Label reachability, IP
       multicast addresses) across areas.

       Distribution tree pruning information is only an optimization, as
       long as multi-destination packets are not prematurely pruned.
       For instance, border TRILL switches could advertise they can
       reach all possible Data Labels, and have an IP multicast router
       attached.  This would cause all multi-destination traffic to be
       transmitted to border TRILL switches, and possibly pruned there,
       when the traffic could have been pruned earlier based on Data
       Label or multicast group if border TRILL switches advertised more
       detailed Data Label and/or multicast listener and multicast
       router attachment information.

   d.  Computation of distribution trees across areas for multi-
       destination data.

       See Section 2.3.

   e.  Computation of RPF information for those distribution trees.

       See Section 2.4.

   f.  Computation of pruning information across areas.

       See Sections 2.3 and 2.6.










Perlman, et al.               Informational                    [Page 10]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   g.  Compatibility, as much as practical, with existing, unmodified
       TRILL switches.

       The most important form of compatibility is with existing TRILL
       fast-path hardware.  Changes that require upgrade to the slow-
       path firmware/software are more tolerable.  Compatibility for the
       relatively small number of border TRILL switches is less
       important than compatibility for non-border TRILL switches.

       See Section 5.

2.1.  Non-Zero Area Addresses



   The current TRILL base protocol specification [RFC6325] [RFC7177]
   [RFC7780] says that the area address in IS-IS must be zero.  The
   purpose of the area address is to ensure that different areas are not
   accidentally merged.  Furthermore, zero is an invalid area address
   for Layer 3 IS-IS, so it was chosen as an additional safety mechanism
   to ensure that Layer 3 IS-IS packets would not be confused with TRILL
   IS-IS packets.  However, TRILL uses other techniques to avoid
   confusion on a link, such as different multicast addresses and
   Ethertypes on Ethernet [RFC6325], different PPP (Point-to-Point
   Protocol) code points on PPP [RFC6361], and the like.  Thus, using an
   area address in TRILL that might be used in Layer 3 IS-IS is not a
   problem.

   Since current TRILL switches will reject any IS-IS messages with non-
   zero area addresses, the choices are as follows:

   a.1.  upgrade all TRILL switches that are to interoperate in a
         potentially multilevel environment to understand non-zero area
         addresses,

   a.2.  neighbors of old TRILL switches must remove the area address
         from IS-IS messages when talking to an old TRILL switch (which
         might break IS-IS security and/or cause inadvertent merging of
         areas),

   a.3.  ignore the problem of accidentally merging areas entirely, or

   a.4.  keep the fixed "area address" field as 0 in TRILL, and add a
         new, optional TLV for "area name" to Hellos that, if present,
         could be compared, by new TRILL switches, to prevent accidental
         area merging.

   In principal, different solutions could be used in different areas
   but it would be much simpler to adopt one of these choices uniformly.
   A simple solution would be a.1, with each TRILL switch using a



Perlman, et al.               Informational                    [Page 11]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   dominant area nickname as its area address.  For the unique-nickname
   alternative, the dominant nickname could be the lowest value nickname
   held by any border RBridge of the area.  For the aggregated-nickname
   alternative, it could be the lowest nickname held by a border RBridge
   of the area or a nickname representing the area.

2.2.  Aggregated versus Unique Nicknames



   In the unique-nickname alternative, all nicknames across the campus
   must be unique.  In the aggregated-nickname alternative, TRILL switch
   nicknames within an aggregated-nickname area are only of local
   significance, and the only nickname externally (outside that area)
   visible is the "area nickname" (or nicknames), which aggregates all
   the internal nicknames.

   The unique-nickname approach simplifies border TRILL switches.

   The aggregated-nickname approach eliminates the potential problem of
   nickname exhaustion, minimizes the amount of nickname information
   that would need to be forwarded between areas, minimizes the size of
   the forwarding table, and simplifies RPF calculation and RPF
   information.

2.2.1.  More Details on Unique Nicknames



   With unique cross-area nicknames, it would be intractable to have a
   flat nickname space with TRILL switches in different areas contending
   for the same nicknames.  Instead, each area would need to be
   configured with or allocate one or more blocks of nicknames.  Either
   some TRILL switches would need to announce that all the nicknames
   other than those in blocks available to the area are taken (to
   prevent the TRILL switches inside the area from choosing nicknames
   outside the area's nickname block) or a new TLV would be needed to
   announce the allowable or the prohibited nicknames, and all TRILL
   switches in the area would need to understand that new TLV.

   Currently, the encoding of nickname information in TLVs is by listing
   of individual nicknames; this would make it painful for a border
   TRILL switch to announce into an area that it is holding all other
   nicknames to limit the nicknames available within that area.  Painful
   means tens of thousands of individual nickname entries in the Level 1
   LSDB.  The information could be encoded as ranges of nicknames to
   make this manageable by specifying a new TLV similar to the Nickname
   Flags APPsub-TLV specified in [RFC7780] but providing flags for
   blocks of nicknames rather than single nicknames.  Although this
   would require updating software, such a new TLV is the preferred
   method.




Perlman, et al.               Informational                    [Page 12]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   There is also an issue with the unique-nickname approach in building
   distribution trees, as follows:

      With unique nicknames in the TRILL campus and TRILL header
      nicknames not rewritten by the border TRILL switches, there would
      have to be globally known nicknames for the trees.  Suppose there
      are k trees.  For all of the trees with nicknames located outside
      an area, the local trees would be rooted at a border TRILL switch
      or switches.  Therefore, there would be either no splitting of
      multi-destination traffic within the area or restricted splitting
      of multi-destination traffic between trees rooted at a highly
      restricted set of TRILL switches.

      As an alternative, just the "egress nickname" field of multi-
      destination TRILL Data packets could be mapped at the border,
      leaving known unicast packets unmapped.  However, this surrenders
      much of the unique nickname advantage of simpler border TRILL
      switches.

   Scaling to a very large campus with unique nicknames might exhaust
   the 16-bit TRILL nicknames space particularly if (1) additional
   nicknames are consumed to support active-active end-station groups at
   the TRILL edge using the techniques standardized in [RFC7781] and (2)
   use of the nickname space is less efficient due to the allocation of,
   for example, power-of-two size blocks of nicknames to areas in the
   same way that use of the IP address space is made less efficient by
   hierarchical allocation (see [RFC3194]).  One method to avoid
   nickname exhaustion might be to expand nicknames to 24 bits; however,
   that technique would require TRILL message format and fast-path
   processing changes and all TRILL switches in the campus to understand
   larger nicknames.


2.2.2.  More Details on Aggregated Nicknames



   The aggregated-nickname approach enables passing far less nickname
   information.  It works as follows, assuming both the source and
   destination areas are using aggregated nicknames:

   There are at least two ways areas could be identified.

      One method would be to assign each area a 16-bit nickname.  This
      would not be the nickname of any actual TRILL switch.  Instead, it
      would be the nickname of the area itself.  Border TRILL switches
      would know the area nickname for their own area(s).  For an
      example of a more-specific multilevel proposal using unique
      nicknames, see [UNIQUE].




Perlman, et al.               Informational                    [Page 13]

RFC 8243              Multilevel TRILL Alternatives       September 2017


      Alternatively, areas could be identified by the set of nicknames
      that identify the border routers for that area.  (See [SingleName]
      for a multilevel proposal using such a set of nicknames.)

   The TRILL Header nickname fields in TRILL Data packets being
   transported through a multilevel TRILL campus with aggregated
   nicknames are as follows:

   -  When both the ingress and egress TRILL switches are in the same
      area, there need be no change from the existing base TRILL
      protocol standard in the TRILL Header nickname fields.

   -  When being transported between different Level 1 areas in Level 2,
      the ingress nickname is a nickname of the ingress TRILL switch's
      area, whereas the egress nickname is either a nickname of the
      egress TRILL switch's area or a tree nickname.

   -  When being transported from Level 1 to Level 2, the ingress
      nickname is the nickname of the ingress TRILL switch itself,
      whereas the egress nickname is either a nickname for the area of
      the egress TRILL switch or a tree nickname.

   -  When being transported from Level 2 to Level 1, the ingress
      nickname is a nickname for the ingress TRILL switch's area,
      whereas the egress nickname is either the nickname of the egress
      TRILL switch itself or a tree nickname.

   There are two variations of the aggregated-nickname approach.  The
   first is the Border Learning approach, which is described in
   Section 2.2.2.1.  The second is the Swap Nickname Field approach,
   which is described in Section 2.2.2.2.  Section 2.2.2.3 compares the
   advantages and disadvantages of these two variations of the
   aggregated-nickname approach.

2.2.2.1.  Border Learning Aggregated Nicknames



   This section provides an illustrative example and description of the
   border-learning variation of aggregated nicknames where a single
   nickname is used to identify an area.

   In the following picture, RB2 and RB3 are area border TRILL switches
   (RBridges).  A source S is attached to RB1.  The two areas have
   nicknames 15961 and 15918, respectively.  RB1 has a nickname, say 27,
   and RB4 has a nickname, say 44 (and in fact, they could even have the
   same nickname, since the TRILL switch nickname will not be visible
   outside these aggregated-nickname areas).





Perlman, et al.               Informational                    [Page 14]

RFC 8243              Multilevel TRILL Alternatives       September 2017


            Area 15961              level 2             Area 15918
    +-------------------+     +-----------------+     +--------------+
    |                   |     |                 |     |              |
    |  S--RB1---Rx--Rz----RB2---Rb---Rc--Rd---Re--RB3---Rk--RB4---D  |
    |     27            |     |                 |     |     44       |
    |                   |     |                 |     |              |
    +-------------------+     +-----------------+     +--------------+

   Let's say that S transmits a frame to destination D, which is
   connected to RB4, and let's say that D's location has already been
   learned by the relevant TRILL switches.  These relevant switches have
   learned the following:

   1.  RB1 has learned that D is connected to nickname 15918
   2.  RB3 has learned that D is attached to nickname 44.

   The following sequence of events will occur:

   -  S transmits an Ethernet frame with source MAC = S and destination
      MAC = D.

   -  RB1 encapsulates with a TRILL header with ingress RBridge = 27,
      and egress = 15918 producing a TRILL Data packet.

   -  RB2 has announced in the Level 1 IS-IS instance in area 15961,
      that it is attached to all the area nicknames, including 15918.
      Therefore, IS-IS routes the packet to RB2.  Alternatively, if a
      distinguished range of nicknames is used for Level 2, Level 1
      TRILL switches seeing such an egress nickname will know to route
      to the nearest border router, which can be indicated by the IS-IS
      "attached bit" [IS-IS].

   -  RB2, when transitioning the packet from Level 1 to Level 2,
      replaces the ingress TRILL switch nickname with the area nickname,
      so it replaces 27 with 15961.  Within Level 2, the ingress RBridge
      field in the TRILL header will therefore be 15961, and the egress
      RBridge field will be 15918.  Also RB2 learns that S is attached
      to nickname 27 in area 15961 to accommodate return traffic.

   -  The packet is forwarded through Level 2, to RB3, which has
      advertised, in Level 2, reachability to the nickname 15918.

   -  RB3, when forwarding into area 15918, replaces the egress nickname
      in the TRILL header with RB4's nickname (44).  So, within the
      destination area, the ingress nickname will be 15961 and the
      egress nickname will be 44.





Perlman, et al.               Informational                    [Page 15]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   -  RB4, when decapsulating, learns that S is attached to nickname
      15961, which is the area nickname of the ingress.

   Now suppose that D's location has not been learned by RB1 and/or RB3.
   What will happen, as it would in TRILL today, is that RB1 will
   forward the packet as multi-destination, choosing a tree.  As the
   multi-destination packet transitions into Level 2, RB2 replaces the
   ingress nickname with the area nickname.  If RB1 does not know the
   location of D, the packet must be flooded, subject to possible
   pruning, in Level 2 and, subject to possible pruning, from Level 2
   into every Level 1 area that it reaches on the Level 2 distribution
   tree.

   Now suppose that RB1 has learned the location of D (attached to
   nickname 15918), but RB3 does not know where D is.  In that case, RB3
   must turn the packet into a multi-destination packet within area
   15918.  In this case, care must be taken so that in the case in which
   RB3 is not the designated transitioner between Level 2 and its area
   for that multi-destination packet, but was on the unicast path, that
   border TRILL switch in that area does not forward the now multi-
   destination packet back into Level 2.  Therefore, it would be
   desirable to have a marking, somehow, that indicates the scope of
   this packet's distribution to be "only this area" (see also
   Section 4).

   In cases where there are multiple transitioners for unicast packets,
   the border-learning mode of operation requires that the address
   learning between them be shared by some protocol such as running
   ESADI [RFC7357] for all Data Labels of interest to avoid excessive
   unknown unicast flooding.

   The potential issue described at the end of Section 2.2.1 with trees
   in the unique-nickname alternative is eliminated with aggregated
   nicknames.  With aggregated nicknames, each border TRILL switch that
   will transition multi-destination packets can have a mapping between
   Level 2 tree nicknames and Level 1 tree nicknames.  There need not
   even be agreement about the total number of trees: just agreement
   that the border TRILL switch have some mapping and replace the egress
   TRILL switch nickname (the tree name) when transitioning levels.












Perlman, et al.               Informational                    [Page 16]

RFC 8243              Multilevel TRILL Alternatives       September 2017


2.2.2.2.  Swap Nickname Field Aggregated Nicknames



   There is a variant possibility where two additional fields could
   exist in TRILL Data packets that could be called the "ingress swap
   nickname field" and the "egress swap nickname field".  This variant
   is described below for completeness, but it would require fast-path
   hardware changes from the existing TRILL protocol.  The changes in
   the example above would be as follows:

   -  RB1 will have learned the area nickname of D and the TRILL switch
      nickname of RB4 to which D is attached.  In encapsulating a frame
      to D, it puts an area nickname of D (15918) in the egress nickname
      field of the TRILL Header and puts a nickname of RB3 (44) in an
      egress swap nickname field.

   -  RB2 moves the ingress nickname to the ingress swap nickname field
      and inserts 15961, an area nickname for S, into the ingress
      nickname field.

   -  RB3 swaps the egress nickname and the egress swap nickname fields,
      which sets the egress nickname to 44.

   -  RB4 learns the correspondence between the source MAC/VLAN of S and
      the { ingress nickname, ingress swap nickname field } pair as it
      decapsulates and egresses the frame.

   See [TRILL-IP] for a multilevel proposal using aggregated swap
   nicknames with a single nickname representing an area.

2.2.2.3.  Comparison



   The border-learning variant described in Section 2.2.2.1 minimizes
   the change in non-border TRILL switches, but it imposes the burden on
   border TRILL switches of learning and doing lookups in all the end-
   station MAC addresses within their area(s) that are used for
   communication outside the area.  This burden could be reduced by
   decreasing the area size and increasing the number of areas.

   The Swap Nickname Field variant described in Section 2.2.2.2
   eliminates the extra address learning burden on border TRILL
   switches, but it requires changes to the TRILL Data packet header and
   more extensive changes to non-border TRILL switches.  In particular,
   with this alternative, non-border TRILL switches must learn to
   associate both a TRILL switch nickname and an area nickname with end-
   station MAC/label pairs (except for addresses that are local to their
   area).





Perlman, et al.               Informational                    [Page 17]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   The Swap Nickname Field alternative is more scalable but less
   backward compatible for non-border TRILL switches.  It would be
   possible for border and other Level 2 TRILL switches to support both
   border learning, for support of legacy Level 1 TRILL switches, and
   Swap Nickname Field, to support Level 1 TRILL switches that
   understood the Swap Nickname Field method based on variations in the
   TRILL header; however, this would be even more complex.

   The requirement to change the TRILL header and fast-path processing
   to support the Swap Nickname Field variant make it impractical for
   the foreseeable future.

2.3.  Building Multi-Area Trees



   It is easy to build a multi-area tree by building a tree in each area
   separately, (including the Level 2 area), and then having only a
   single-border TRILL switch, say RBx, in each area, attach to the
   Level 2 area.  RBx would forward all multi-destination packets
   between that area and Level 2.

   However, people might find this unacceptable because of the desire to
   path split (not always sending all multi-destination traffic through
   the same border TRILL switch).

   This is the same issue as with multiple ingress TRILL switches
   injecting traffic from a pseudonode, and it can be solved with the
   mechanism that was adopted for that purpose: the affinity TLV
   [RFC7783].  For each tree in the area, at most one border RB
   announces itself in an affinity TLV with that tree name.

2.4.  The RPF Check for Trees



   For multi-destination data originating locally in RBx's area,
   computation of the RPF check is done as today.  For multi-destination
   packets originating outside RBx's area, computation of the RPF check
   must be done based on which one of the border TRILL switches (say
   RB1, RB2, or RB3) injected the packet into the area.

   A TRILL switch, say RB4, located inside an area, must be able to know
   which of RB1, RB2, or RB3 transitioned the packet into the area from
   Level 2 (or into Level 2 from an area).

   This could be done using any one of a variety of mechanisms such as
   having the DBRB announce the transitioner assignments to all the
   TRILL switches in the area or using the Affinity sub-TLV mechanism
   given in [RFC7783] or with a New Tree Encoding mechanism discussed in
   Section 4.1.1.




Perlman, et al.               Informational                    [Page 18]

RFC 8243              Multilevel TRILL Alternatives       September 2017


2.5.  Area Nickname Acquisition



   In the aggregated-nickname alternative, each area must acquire a
   unique identifier, for example, by acquiring a unique area nickname
   or by using an identifier based on the area's set of border TRILL
   switches.  It is probably simpler to allocate a block of nicknames
   (say, the top 4000) to either (1) represent areas and not specific
   TRILL switches or (2) be used by border TRILL switches if the set of
   such border TRILL switches represent the area.

   The nicknames used for area identification need to be advertised and
   acquired through Level 2.

   Within an area, all the border TRILL switches can discover each other
   through the Level 1 LSDB, by using the IS-IS "attached bit" [IS-IS]
   or by explicitly advertising in their LSP "I am a border RBridge".

   Of the border TRILL switches, one will have highest priority (say
   RB7).  RB7 can dynamically participate, in Level 2, to acquire a
   nickname for identifying the area.  Alternatively, RB7 could give the
   area a pseudonode IS-IS ID, such as RB7.5, within Level 2.  So an
   area would appear, in Level 2, as a pseudonode and the pseudonode
   could participate, in Level 2, to acquire a nickname for the area.

   Within Level 2, all the border TRILL switches for an area can
   advertise reachability to the area, which would mean connectivity to
   a nickname identifying the area.

2.6.  Link State Representation of Areas



   Within an area, say area A1, there is an election for the DBRB, say
   RB1.  This can be done through LSPs within area A1.  The border TRILL
   switches announce themselves, together with their DBRB priority.
   (Note that the election of the DBRB cannot be done based on Hello
   messages, because the border TRILL switches are not necessarily
   physical neighbors of each other.  They can, however, reach each
   other through connectivity within the area, which is why it will work
   to find each other through Level 1 LSPs.)

   RB1 can acquire an area nickname (in the aggregated-nickname
   approach), and may give the area a pseudonode IS-IS ID (just like the
   Designated RBridge (DRB) would give a pseudonode IS-IS ID to a link)
   depending on how the area nickname is handled.  RB1 advertises, in
   area A1, an area nickname that RB1 has acquired (and what the
   pseudonode IS-IS ID for the area is if needed).






Perlman, et al.               Informational                    [Page 19]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   Level 1 LSPs (possibly pseudonode) initiated by RB1 for the area
   include any information external to area A1 that should be input into
   area A1 (such as nicknames of external areas, or perhaps (in the
   unique nickname variant) all the nicknames of external TRILL switches
   in the TRILL campus and pruning information such as multicast
   listeners and labels).  All the other border TRILL switches for the
   area announce (in their LSP) attachment to that area.

   Within Level 2, RB1 generates a Level 2 LSP on behalf of the area.
   The same pseudonode ID could be used within Level 1 and Level 2, for
   the area.  (There does not seem any reason why it would be useful for
   it to be different, but there's also no reason why it would need to
   be the same).  Likewise, all the area A1 border TRILL switches would
   announce, in their Level 2 LSPs, connection to the area.

3.  Area Partition

   It is possible for an area to become partitioned, so that there is
   still a path from one section of the area to the other, but that path
   is via the Level 2 area.

   With multilevel TRILL, an area will naturally break into two areas in
   this case.

   Area addresses might be configured to ensure two areas are not
   inadvertently connected.  Area addresses appear in Hellos and LSPs
   within the area.  If two chunks, connected only via Level 2, were
   configured with the same area address, this would not cause any
   problems.  (They would just operate as separate Level 1 areas.)

   A more serious problem occurs if the Level 2 area is partitioned in
   such a way that it could be healed by using a path through a Level 1
   area.  TRILL will not attempt to solve this problem.  Within the
   Level 1 area, a single-border RBridge will be the DBRB, and will be
   in charge of deciding which (single) RBridge will transition any
   particular multi-destination packets between that area and Level 2.
   If the Level 2 area is partitioned, this will result in multi-
   destination data only reaching the portion of the TRILL campus
   reachable through the partition attached to the TRILL switch that
   transitions that packet.  It will not cause a loop.











Perlman, et al.               Informational                    [Page 20]

RFC 8243              Multilevel TRILL Alternatives       September 2017


4.  Multi-Destination Scope

   There are at least two reasons it would be desirable to be able to
   mark a multi-destination packet with a scope that indicates the
   packet should not exit the area, as follows:

   1.  To address an issue in the border learning variant of the
       aggregated-nickname alternative, when a unicast packet turns into
       a multi-destination packet when transitioning from Level 2 to
       Level 1, as discussed in Section 4.1.

   2.  To constrain the broadcast domain for certain discovery,
       directory, or service protocols as discussed in Section 4.2.

   Multi-destination packet distribution scope restriction could be done
   in a number of ways.  For example, there could be a flag in the
   packet that means "for this area only".  However, the technique that
   might require the least change to TRILL switch fast-path logic would
   be to indicate this in the egress nickname that designates the
   distribution tree being used.  There could be two general tree
   nicknames for each tree, one being for distribution restricted to the
   area and the other being for multi-area trees.  Or there would be a
   set of N (perhaps 16) special currently reserved nicknames used to
   specify the N highest priority trees but with the variation that if
   the special nickname is used for the tree, the packet is not
   transitioned between areas.  Or one or more special trees could be
   built that were restricted to the local area.

4.1.  Unicast to Multi-Destination Conversions



   In the border learning variant of the aggregated-nickname
   alternative, the following situation may occur:

   -  a unicast packet might be known at the Level 1 to Level 2
      transition and be forwarded as a unicast packet to the least-cost
      border TRILL switch advertising connectivity to the destination
      area, but

   -  upon arriving at the border TRILL switch, it turns out to have an
      unknown destination { MAC, Data Label } pair.

   In this case, the packet must be converted into a multi-destination
   packet and flooded in the destination area.  However, if the border
   TRILL switch doing the conversion is not the border TRILL switch
   designated to transition the resulting multi-destination packet,
   there is the danger that the designated transitioner may pick up the
   packet and flood it back into Level 2 from which it may be flooded
   into multiple areas.  This danger can be avoided by restricting any



Perlman, et al.               Informational                    [Page 21]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   multi-destination packet that results from such a conversion to the
   destination area as described above.

   Alternatively, a multi-destination packet intended only for the area
   could be tunneled (within the area) to the RBridge RBx, that is the
   appointed transitioner for that form of packet (say, based on VLAN or
   FGL), with instructions that RBx only transmit the packet within the
   area, and RBx could initiate the multi-destination packet within the
   area.  Since RBx introduced the packet, and is the only one allowed
   to transition that packet to Level 2, this would accomplish scoping
   of the packet to within the area.  Since this case only occurs in the
   unusual case when unicast packets need to be turned into multi-
   destination as described above, the suboptimality of tunneling
   between the border TRILL switch that receives the unicast packet and
   the appointed level transitioner for that packet might not be an
   issue.

4.1.1.  New Tree Encoding



   The current encoding, in a TRILL header of a tree, is of the nickname
   of the tree root.  This requires all 16 bits of the egress nickname
   field.  TRILL could instead, for example, use the bottom 6 bits to
   encode the tree number (allowing 64 trees), leaving 10 bits to encode
   information such as:

   scope:  a flag indicating whether it should be single area only or an
      entire campus

   border injector:  an indicator of which of the k border TRILL
      switches injected this packet

   If TRILL were to adopt this new encoding, any of the TRILL switches
   in an edge group could inject a multi-destination packet.  This would
   require all TRILL switches to be changed to understand the new
   encoding for a tree, and it would require a TLV in the LSP to
   indicate which number each of the TRILL switches in an edge group
   would be.

   While there are a number of advantages to this technique, it requires
   fast-path logic changes; thus, its deployment is not practical at
   this time.  It is included here for completeness.

4.2.  Selective Broadcast Domain Reduction



   There are a number of service, discovery, and directory protocols
   that, for convenience, are accessed via multicast or broadcast
   frames.  Examples are DHCP, the NetBIOS Service Location Protocol,
   and multicast DNS.



Perlman, et al.               Informational                    [Page 22]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   Some such protocols provide means to restrict distribution to an IP
   subnet or equivalent to reduce size of the broadcast domain they are
   using, and then they provide a proxy that can be placed in that
   subnet to use unicast to access a service elsewhere.  In cases where
   a proxy mechanism is not currently defined, it may be possible to
   create one that references a central server or cache.  With
   multilevel TRILL, it is possible to construct very large IP subnets
   that could become saturated with multi-destination traffic of this
   type unless packets can be further restricted in their distribution.
   Such restricted distribution can be accomplished for some protocols,
   say protocol P, in a variety of ways including the following:

   -  Either (1) at all ingress TRILL switches in an area, place all
      protocol P multi-destination packets on a distribution tree in
      such a way that the packets are restricted to the area or (2) at
      all border TRILL switches between that area and Level 2, detect
      protocol P multi-destination packets and do not transition them.

   -  Then, place one, or a few for redundancy, protocol P proxies
      inside each area where protocol P may be in use.  These proxies
      unicast protocol P requests or other messages to the actual campus
      server(s) for P.  They also receive unicast responses or other
      messages from those servers and deliver them within the area via
      unicast, multicast, or broadcast as appropriate.  (Such proxies
      would not be needed if it was acceptable for all protocol P
      traffic to be restricted to an area.)

   While it might seem logical to connect the campus servers to TRILL
   switches in Level 2, they could be placed within one or more areas so
   that, in some cases, those areas might not require a local proxy
   server.

5.  Coexistence with Old TRILL Switches



   TRILL switches that are not multilevel aware may have a problem with
   calculating RPF check and filtering information, since they would not
   be aware of the assignment of border TRILL switch transitioning.

   A possible solution, as long as any old TRILL switches exist within
   an area, is to have the border TRILL switches elect a single DBRB and
   have all inter-area traffic go through the DBRB (unicast as well as
   multi-destination).  If that DBRB goes down, a new one will be
   elected, but at any one time, all inter-area traffic (unicast as well
   as multi-destination) would go through that one DRBR.  However this
   eliminates load splitting at level transition.






Perlman, et al.               Informational                    [Page 23]

RFC 8243              Multilevel TRILL Alternatives       September 2017


6.  Multi-Access Links with End Stations

   Care must be taken in the case where there are multiple TRILL
   switches on a link with one or more end stations, keeping in mind
   that end stations are TRILL ignorant.  In particular, it is essential
   that only one TRILL switch ingress/egress any given data packet from/
   to an end station so that connectivity is provided to that end
   station without duplicating end-station data and that loops are not
   formed due to one TRILL switch egressing data in native form (i.e.,
   with no TRILL header) and having that data re-ingressed by another
   TRILL switch on the link.

   With existing, single-level TRILL, this is done by electing a single
   DRB per link, which appoints a single Appointed Forwarder per VLAN
   [RFC7177] [RFC8139].  This mechanism depends on the RBridges
   establishing adjacency.  But, suppose there are two (or more) TRILL
   switches on a link in different areas, say RB1 in area A1 and RB2 in
   area A2, as shown below; and suppose that the link also has one or
   more end stations attached.  If RB1 and RB2 ignore each other's
   Hellos because they are in different areas, as they are required to
   do under normal IS-IS PDU processing rules, then they will not form
   an adjacency.  If they are not adjacent, they will ignore each other
   for the Appointed Forwarder mechanism and will both ingress/egress
   end-station traffic on the link causing loops and duplication.

   The problem is not avoiding adjacency or avoiding TRILL-Data-packet
   transfer between RB1 and RB2; the area address mechanism of IS-IS or
   possibly the use of topology constraints (or the like) does that
   quite well.  The problem stems from end stations being TRILL
   ignorant; therefore, care must be taken so that multiple RBridges on
   a link do not ingress the same frame originated by an end station and
   so that an RBridge does not ingress a native frame egressed by a
   different RBridge because the RBridge mistakes the frame for a frame
   originated by an end station.

















Perlman, et al.               Informational                    [Page 24]

RFC 8243              Multilevel TRILL Alternatives       September 2017


      +--------------------------------------------+
      |                   Level 2                  |
      +----------+---------------------+-----------+
      |  Area A1 |                     |  Area A2  |
      |   +---+  |                     |   +---+   |
      |   |RB1|  |                     |   |RB2|   |
      |   +-+-+  |                     |   +-+-+   |
      |     |    |                     |     |     |
      +-----|----+                     +-----|-----+
            |                                |
          --+---------+-------------+--------+-- Link
                      |             |
               +------+------+   +--+----------+
               | End Station |   | End Station |
               +-------------+   +-------------+

   A simple rule, which is preferred, is to use the TRILL switch or
   switches having the lowest-numbered area, comparing area numbers as
   unsigned integers, to handle all native traffic to/from end stations
   on the link.  This would automatically give multilevel-ignorant
   legacy TRILL switches, that would be using area number zero, highest
   priority for handling end-station traffic, which they would try to do
   anyway.

   Other methods are possible.  For example, making the selection of the
   Appointed Forwarders and the TRILL switch in charge of that selection
   across all TRILL switches on the link, regardless of area.  However,
   a special case would then have to be made for legacy TRILL switches
   using area number zero.

   These techniques require multilevel-aware TRILL switches to take
   actions based on Hellos from RBridges in other areas, even though
   they will not form an adjacency with such RBridges.  However, the
   action is quite simple in the preferred case: if a TRILL switch sees
   Hellos from lower-numbered areas, then they would not act as an
   Appointed Forwarder on the link until the Hello timer for such Hellos
   had expired.

7.  Summary

   This document describes potential scaling issues in TRILL and
   possible approaches to multilevel TRILL as a solution or element of a
   solution to most of them.

   The alternative using aggregated-nickname areas in multilevel TRILL
   has significant advantages in terms of scalability over using campus-
   wide unique nicknames, not just in avoiding nickname exhaustion, but
   by allowing RPF checks to be aggregated based on an entire area.



Perlman, et al.               Informational                    [Page 25]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   However, the alternative of using unique nicknames is simpler and
   avoids the changes in border TRILL switches required to support
   aggregated nicknames.  It is possible to support both.  For example,
   a TRILL campus could use simpler unique nicknames until scaling
   begins to cause problems and then start to introduce areas with
   aggregated nicknames.

   Some multilevel TRILL issues are not difficult, such as dealing with
   partitioned areas.  Other issues are more difficult, especially
   dealing with old TRILL switches that are multilevel ignorant.

8.  Security Considerations



   This informational document explores alternatives for the design of
   multilevel IS-IS in TRILL and generally does not consider security
   issues.

   If aggregated nicknames are used in two areas that have the same area
   address, and those areas merge, there is a possibility of a transient
   nickname collision that would not occur with unique nicknames.  Such
   a collision could cause a data packet to be delivered to the wrong
   egress TRILL switch, but it would still not be delivered to any end
   station in the wrong Data Label; thus, such delivery would still
   conform to security policies.

   For general TRILL security considerations, see [RFC6325].

9.  IANA Considerations



   This document does not require any IANA actions.

10.  References



10.1.  Normative References



   [IS-IS]    International Organization for Standardization,
              "Information technology -- Telecommunications and
              information exchange between systems -- Intermediate
              System to Intermediate System intra-domain routeing
              information exchange protocol for use in conjunction with
              the protocol for providing the connectionless-mode network
              service (ISO 8473)", ISO/IEC 10589:2002, Second Edition,
              November 2002.

   [RFC6325]  Perlman, R., Eastlake 3rd, D., Dutt, D., Gai, S., and A.
              Ghanwani, "Routing Bridges (RBridges): Base Protocol
              Specification", RFC 6325, DOI 10.17487/RFC6325, July 2011,
              <https://www.rfc-editor.org/info/rfc6325>.



Perlman, et al.               Informational                    [Page 26]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   [RFC7177]  Eastlake 3rd, D., Perlman, R., Ghanwani, A., Yang, H., and
              V. Manral, "Transparent Interconnection of Lots of Links
              (TRILL): Adjacency", RFC 7177, DOI 10.17487/RFC7177, May
              2014, <https://www.rfc-editor.org/info/rfc7177>.

   [RFC7780]  Eastlake 3rd, D., Zhang, M., Perlman, R., Banerjee, A.,
              Ghanwani, A., and S. Gupta, "Transparent Interconnection
              of Lots of Links (TRILL): Clarifications, Corrections, and
              Updates", RFC 7780, DOI 10.17487/RFC7780, February 2016,
              <https://www.rfc-editor.org/info/rfc7780>.

   [RFC8139]  Eastlake 3rd, D., Li, Y., Umair, M., Banerjee, A., and F.
              Hu, "Transparent Interconnection of Lots of Links (TRILL):
              Appointed Forwarders", RFC 8139, DOI 10.17487/RFC8139,
              June 2017, <https://www.rfc-editor.org/info/rfc8139>.

10.2.  Informative References



   [RFC3194]  Durand, A. and C. Huitema, "The H-Density Ratio for
              Address Assignment Efficiency An Update on the H ratio",
              RFC 3194, DOI 10.17487/RFC3194, November 2001,
              <https://www.rfc-editor.org/info/rfc3194>.

   [RFC6361]  Carlson, J. and D. Eastlake 3rd, "PPP Transparent
              Interconnection of Lots of Links (TRILL) Protocol Control
              Protocol", RFC 6361, DOI 10.17487/RFC6361, August 2011,
              <https://www.rfc-editor.org/info/rfc6361>.

   [RFC7172]  Eastlake 3rd, D., Zhang, M., Agarwal, P., Perlman, R., and
              D. Dutt, "Transparent Interconnection of Lots of Links
              (TRILL): Fine-Grained Labeling", RFC 7172,
              DOI 10.17487/RFC7172, May 2014,
              <https://www.rfc-editor.org/info/rfc7172>.

   [RFC7176]  Eastlake 3rd, D., Senevirathne, T., Ghanwani, A., Dutt,
              D., and A. Banerjee, "Transparent Interconnection of Lots
              of Links (TRILL) Use of IS-IS", RFC 7176,
              DOI 10.17487/RFC7176, May 2014,
              <https://www.rfc-editor.org/info/rfc7176>.

   [RFC7357]  Zhai, H., Hu, F., Perlman, R., Eastlake 3rd, D., and O.
              Stokes, "Transparent Interconnection of Lots of Links
              (TRILL): End Station Address Distribution Information
              (ESADI) Protocol", RFC 7357, DOI 10.17487/RFC7357,
              September 2014, <https://www.rfc-editor.org/info/rfc7357>.






Perlman, et al.               Informational                    [Page 27]

RFC 8243              Multilevel TRILL Alternatives       September 2017


   [RFC7781]  Zhai, H., Senevirathne, T., Perlman, R., Zhang, M., and Y.
              Li, "Transparent Interconnection of Lots of Links (TRILL):
              Pseudo-Nickname for Active-Active Access", RFC 7781,
              DOI 10.17487/RFC7781, February 2016,
              <https://www.rfc-editor.org/info/rfc7781>.

   [RFC7783]  Senevirathne, T., Pathangi, J., and J. Hudson,
              "Coordinated Multicast Trees (CMT) for Transparent
              Interconnection of Lots of Links (TRILL)", RFC 7783,
              DOI 10.17487/RFC7783, February 2016,
              <https://www.rfc-editor.org/info/rfc7783>.

   [InterCon] Perlman, R., "Interconnection: Bridges, Routers, Switches,
              and Internetworking Protocols", Addison Wesley
              Longman, Second Edition, Chapter 3, 1999.

   [TRILL-IP] Bhikkaji, B., Venkataswami, B., Mahadevan, R., Sundaram,
              S., and N. Swamy, "Connecting Disparate Data Center/PBB/
              Campus TRILL sites using BGP", Work in Progress,
              draft-balaji-trill-over-ip-multi-level-05, March 2012.

   [UNIQUE]   Zhang, M., Eastlake, D., Perlman, R., Cullen, M., Zhai,
              H., and D. Liu, "TRILL Multilevel Using Unique Nicknames",
              Work in Progress, draft-ietf-trill-multilevel-
              unique-nickname-02, May 2017.

   [SingleName]
              Zhang, M., Eastlake, D., Perlman, R., Cullen, M., and H.
              Zhai, "Transparent Interconnection of Lots of Links
              (TRILL) Single Area Border RBridge Nickname for
              Multilevel", Work in Progress, draft-ietf-trill-
              multilevel-single-nickname-04, July 2017.

Acknowledgements



   The helpful comments and contributions of the following are hereby
   acknowledged:

      Alia Atlas, David Michael Bond, Dino Farinacci, Sue Hares, Gayle
      Noble, Alexander Vainshtein, and Stig Venaas.











Perlman, et al.               Informational                    [Page 28]

RFC 8243              Multilevel TRILL Alternatives       September 2017


Authors' Addresses



   Radia Perlman
   EMC
   2010 256th Avenue NE, #200
   Bellevue, WA 98007
   United States of America

   Email: radia@alum.mit.edu


   Donald Eastlake 3rd
   Huawei Technologies
   155 Beaver Street
   Milford, MA 01757
   United States of America

   Phone: +1-508-333-2270
   Email: d3e3e3@gmail.com


   Mingui Zhang
   Huawei Technologies
   No.156 Beiqing Rd. Haidian District,
   Beijing 100095
   China

   Email: zhangmingui@huawei.com


   Anoop Ghanwani
   Dell
   5450 Great America Parkway
   Santa Clara, CA  95054
   United States of America

   Email: anoop@alumni.duke.edu


   Hongjun Zhai
   Jinling Institute of Technology
   99 Hongjing Avenue, Jiangning District
   Nanjing, Jiangsu 211169
   China

   Email: honjun.zhai@tom.com





Perlman, et al.               Informational                    [Page 29]