Internet Draft
Network Working Group Eric C. Rosen
Internet Draft Cisco Systems, Inc.
Expiration Date: February 1998
Arun Viswanathan
IBM Corp.
Ross Callon
Ascend Communications, Inc.
August 1997
A Proposed Architecture for MPLS
draft-ietf-mpls-arch-00.txt
Status of this Memo
This document is an Internet-Draft. Internet-Drafts are working
documents of the Internet Engineering Task Force (IETF), its areas,
and its working groups. Note that other groups may also distribute
working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
To learn the current status of any Internet-Draft, please check the
"1id-abstracts.txt" listing contained in the Internet-Drafts Shadow
Directories on ftp.is.co.za (Africa), nic.nordu.net (Europe),
munnari.oz.au (Pacific Rim), ds.internic.net (US East Coast), or
ftp.isi.edu (US West Coast).
Abstract
This internet draft contains a draft protocol architecture for
multiprotocol label switching (MPLS). The proposed architecture is
based on other label switching approaches [2-11] as well as on the
MPLS Framework document [1].
Rosen, Viswanathan & Callon [Page 1]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
Table of Contents
1 Introduction to MPLS ............................... 3
1.1 Overview ........................................... 3
1.2 Terminology ........................................ 5
1.3 Acronyms and Abbreviations ......................... 9
1.4 Acknowledgments .................................... 10
2 Outline of Approach ................................ 10
2.1 Labels ............................................. 10
2.2 Upstream and Downstream LSRs ....................... 11
2.3 Labeled Packet ..................................... 11
2.4 Label Assignment and Distribution; Attributes ...... 11
2.5 Label Distribution Protocol (LDP) .................. 12
2.6 The Label Stack .................................... 12
2.7 The Next Hop Label Forwarding Entry (NHLFE) ........ 13
2.8 Incoming Label Map (ILM) ........................... 13
2.9 Stream-to-NHLFE Map (STN) .......................... 13
2.10 Label Swapping ..................................... 14
2.11 Label Switched Path (LSP), LSP Ingress, LSP Egress . 14
2.12 LSP Next Hop ....................................... 16
2.13 Route Selection .................................... 17
2.14 Time-to-Live (TTL) ................................. 18
2.15 Loop Control ....................................... 19
2.15.1 Loop Prevention .................................... 20
2.15.2 Interworking of Loop Control Options ............... 22
2.16 Merging and Non-Merging LSRs ....................... 23
2.16.1 Stream Merge ....................................... 24
2.16.2 Non-merging LSRs ................................... 24
2.16.3 Labels for Merging and Non-Merging LSRs ............ 25
2.16.4 Merge over ATM ..................................... 26
2.16.4.1 Methods of Eliminating Cell Interleave ............. 26
2.16.4.2 Interoperation: VC Merge, VP Merge, and Non-Merge .. 26
2.17 LSP Control: Egress versus Local ................... 27
2.18 Granularity ........................................ 29
2.19 Tunnels and Hierarchy .............................. 30
2.19.1 Hop-by-Hop Routed Tunnel ........................... 30
2.19.2 Explicitly Routed Tunnel ........................... 30
2.19.3 LSP Tunnels ........................................ 30
2.19.4 Hierarchy: LSP Tunnels within LSPs ................. 31
2.19.5 LDP Peering and Hierarchy .......................... 31
2.20 LDP Transport ...................................... 33
2.21 Label Encodings .................................... 33
2.21.1 MPLS-specific Hardware and/or Software ............. 33
2.21.2 ATM Switches as LSRs ............................... 34
2.21.3 Interoperability among Encoding Techniques ......... 35
2.22 Multicast .......................................... 36
Rosen, Viswanathan & Callon [Page 2]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
3 Some Applications of MPLS .......................... 36
3.1 MPLS and Hop by Hop Routed Traffic ................. 36
3.1.1 Labels for Address Prefixes ........................ 36
3.1.2 Distributing Labels for Address Prefixes ........... 36
3.1.2.1 LDP Peers for a Particular Address Prefix .......... 36
3.1.2.2 Distributing Labels ................................ 37
3.1.3 Using the Hop by Hop path as the LSP ............... 38
3.1.4 LSP Egress and LSP Proxy Egress .................... 38
3.1.5 The POP Label ...................................... 39
3.1.6 Option: Egress-Targeted Label Assignment ........... 40
3.2 MPLS and Explicitly Routed LSPs .................... 41
3.2.1 Explicitly Routed LSP Tunnels: Traffic Engineering . 42
3.3 Label Stacks and Implicit Peering .................. 42
3.4 MPLS and Multi-Path Routing ........................ 43
3.5 LSPs may be Multipoint-to-Point Entities ........... 44
3.6 LSP Tunneling between BGP Border Routers ........... 44
3.7 Other Uses of Hop-by-Hop Routed LSP Tunnels ........ 46
3.8 MPLS and Multicast ................................. 46
4 LDP Procedures ..................................... 47
5 Security Considerations ............................ 47
6 Authors' Addresses ................................. 47
7 References ......................................... 47
Appendix A Why Egress Control is Better ....................... 48
Appendix B Why Local Control is Better ........................ 56
1. Introduction to MPLS
1.1. Overview
In connectionless network layer protocols, as a packet travels from
one router hop to the next, an independent forwarding decision is
made at each hop. Each router analyzes the packet header, and runs a
network layer routing algorithm. The next hop for a packet is chosen
based on the header analysis and the result of running the routing
algorithm.
Packet headers contain considerably more information than is needed
simply to choose the next hop. Choosing the next hop can therefore be
thought of as the composition of two functions. The first function
partitions the entire packet forwarding space into "forwarding
equivalence classes (FECs)". The second maps these FECs to a next
hop. Multiple network layer headers which get mapped into the same
FEC are indistinguishable, as far as the forwarding decision is
concerned. The set of packets belonging to the same FEC, traveling
from a common node, will follow the same path and be forwarded in the
Rosen, Viswanathan & Callon [Page 3]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
same manner (for example, by being placed in a common queue) towards
the destination. This set of packets following the same path,
belonging to the same FEC (and therefore being forwarded in a common
manner) may be referred to as a "stream".
In IP forwarding, multiple packets are typically assigned to the same
Stream by a particular router if there is some address prefix X in
that router's routing tables such that X is the "longest match" for
each packet's destination address.
In MPLS, the mapping from packet headers to stream is performed just
once, as the packet enters the network. The stream to which the
packet is assigned is encoded with a short fixed length value known
as a "label". When a packet is forwarded to its next hop, the label
is sent along with it; that is, the packets are "labeled".
At subsequent hops, there is no further analysis of the network layer
header. Rather, the label is used as an index into a table which
specifies the next hop, and a new label. The old label is replaced
with the new label, and the packet is forwarded to its next hop. This
eliminates the need to perform a longest match computation for each
packet at each hop; the computation can be performed just once.
Some routers analyze a packet's network layer header not merely to
choose the packet's next hop, but also to determine a packet's
"precedence" or "class of service", in order to apply different
discard thresholds or scheduling disciplines to different packets. In
MPLS, this can also be inferred from the label, so that no further
header analysis is needed.
The fact that a packet is assigned to a Stream just once, rather than
at every hop, allows the use of sophisticated forwarding paradigms.
A packet that enters the network at a particular router can be
labeled differently than the same packet entering the network at a
different router, and as a result forwarding decisions that depend on
the ingress point ("policy routing") can be easily made. In fact,
the policy used to assign a packet to a Stream need not have only the
network layer header as input; it may use arbitrary information about
the packet, and/or arbitrary policy information as input. Since this
decouples forwarding from routing, it allows one to use MPLS to
support a large variety of routing policies that are difficult or
impossible to support with just conventional network layer
forwarding.
Similarly, MPLS facilitates the use of explicit routing, without
requiring that each IP packet carry the explicit route. Explicit
routes may be useful to support policy routing and traffic
engineering.
Rosen, Viswanathan & Callon [Page 4]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
MPLS makes use of a routing approach whereby the normal mode of
operation is that L3 routing (e.g., existing IP routing protocols
and/or new IP routing protocols) is used by all nodes to determine
the routed path.
MPLS stands for "Multiprotocol" Label Switching, multiprotocol
because its techniques are applicable to ANY network layer protocol.
In this document, however, we focus on the use of IP as the network
layer protocol.
A router which supports MPLS is known as a "Label Switching Router",
or LSR.
A general discussion of issues related to MPLS is presented in "A
Framework for Multiprotocol Label Switching" [1].
1.2. Terminology
This section gives a general conceptual overview of the terms used in
this document. Some of these terms are more precisely defined in
later sections of the document.
aggregate stream synonym of "stream"
DLCI a label used in Frame Relay networks to
identify frame relay circuits
flow a single instance of an application to
application flow of data (as in the RSVP
and IFMP use of the term "flow")
forwarding equivalence class a group of IP packets which are
forwarded in the same manner (e.g.,
over the same path, with the same
forwarding treatment)
frame merge stream merge, when it is applied to
operation over frame based media, so that
the potential problem of cell interleave
is not an issue.
label a short fixed length physically
contiguous identifier which is used to
identify a stream, usually of local
significance.
Rosen, Viswanathan & Callon [Page 5]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
label information base the database of information containing
label bindings
label swap the basic forwarding operation consisting
of looking up an incoming label to
determine the outgoing label,
encapsulation, port, and other data
handling information.
label swapping a forwarding paradigm allowing
streamlined forwarding of data by using
labels to identify streams of data to be
forwarded.
label switched hop the hop between two MPLS nodes, on which
forwarding is done using labels.
label switched path the path created by the concatenation of
one or more label switched hops, allowing
a packet to be forwarded by swapping
labels from an MPLS node to another MPLS
node.
layer 2 the protocol layer under layer 3 (which
therefore offers the services used by
layer 3). Forwarding, when done by the
swapping of short fixed length labels,
occurs at layer 2 regardless of whether
the label being examined is an ATM
VPI/VCI, a frame relay DLCI, or an MPLS
label.
layer 3 the protocol layer at which IP and its
associated routing protocols operate link
layer synonymous with layer 2
loop detection a method of dealing with loops in which
loops are allowed to be set up, and data
may be transmitted over the loop, but the
loop is later detected and closed
loop prevention a method of dealing with loops in which
data is never transmitted over a loop
label stack an ordered set of labels
Rosen, Viswanathan & Callon [Page 6]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
loop survival a method of dealing with loops in which
data may be transmitted over a loop, but
means are employed to limit the amount of
network resources which may be consumed
by the looping data
label switched path The path through one or more LSRs at one
level of the hierarchy followed by a
stream.
label switching router an MPLS node which is capable of
forwarding native L3 packets
merge point the node at which multiple streams and
switched paths are combined into a single
stream sent over a single path.
Mlabel abbreviation for MPLS label
MPLS core standards the standards which describe the core
MPLS technology
MPLS domain a contiguous set of nodes which operate
MPLS routing and forwarding and which are
also in one Routing or Administrative
Domain
MPLS edge node an MPLS node that connects an MPLS domain
with a node which is outside of the
domain, either because it does not run
MPLS, and/or because it is in a different
domain. Note that if an LSR has a
neighboring host which is not running
MPLS, that that LSR is an MPLS edge node.
MPLS egress node an MPLS edge node in its role in handling
traffic as it leaves an MPLS domain
MPLS ingress node an MPLS edge node in its role in handling
traffic as it enters an MPLS domain
MPLS label a label placed in a short MPLS shim
header used to identify streams
MPLS node a node which is running MPLS. An MPLS
node will be aware of MPLS control
protocols, will operate one or more L3
routing protocols, and will be capable of
Rosen, Viswanathan & Callon [Page 7]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
forwarding packets based on labels. An
MPLS node may optionally be also capable
of forwarding native L3 packets.
MultiProtocol Label Switching an IETF working group and the effort
associated with the working group
network layer synonymous with layer 3
stack synonymous with label stack
stream an aggregate of one or more flows,
treated as one aggregate for the purpose
of forwarding in L2 and/or L3 nodes
(e.g., may be described using a single
label). In many cases a stream may be the
aggregate of a very large number of
flows. Synonymous with "aggregate
stream".
stream merge the merging of several smaller streams
into a larger stream, such that for some
or all of the path the larger stream can
be referred to using a single label.
switched path synonymous with label switched path
virtual circuit a circuit used by a connection-oriented
layer 2 technology such as ATM or Frame
Relay, requiring the maintenance of state
information in layer 2 switches.
VC merge stream merge when it is specifically
applied to VCs, specifically so as to
allow multiple VCs to merge into one
single VC
VP merge stream merge when it is applied to VPs,
specifically so as to allow multiple VPs
to merge into one single VP. In this case
the VCIs need to be unique. This allows
cells from different sources to be
distinguished via the VCI.
VPI/VCI a label used in ATM networks to identify
circuits
Rosen, Viswanathan & Callon [Page 8]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
1.3. Acronyms and Abbreviations
ATM Asynchronous Transfer Mode
BGP Border Gateway Protocol
DLCI Data Link Circuit Identifier
FEC Forwarding Equivalence Class
STN Stream to NHLFE Map
IGP Interior Gateway Protocol
ILM Incoming Label Map
IP Internet Protocol
LIB Label Information Base
LDP Label Distribution Protocol
L2 Layer 2
L3 Layer 3
LSP Label Switched Path
LSR Label Switching Router
MPLS MultiProtocol Label Switching
MPT Multipoint to Point Tree
NHLFE Next Hop Label Forwarding Entry
SVC Switched Virtual Circuit
SVP Switched Virtual Path
TTL Time-To-Live
VC Virtual Circuit
VCI Virtual Circuit Identifier
VP Virtual Path
Rosen, Viswanathan & Callon [Page 9]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
VPI Virtual Path Identifier
1.4. Acknowledgments
The ideas and text in this document have been collected from a number
of sources and comments received. We would like to thank Rick Boivie,
Paul Doolan, Nancy Feldman, Yakov Rekhter, Vijay Srinivasan, and
George Swallow for their inputs and ideas.
2. Outline of Approach
In this section, we introduce some of the basic concepts of MPLS and
describe the general approach to be used.
2.1. Labels
A label is a short fixed length locally significant identifier which
is used to identify a stream. The label is based on the stream or
forwarding equivalence class that a packet is assigned to. The label
does not directly encode the network layer address, and is based on
the network layer address only to the extent that the forwarding
equivalence class is based on the address.
If Ru and Rd are neighboring LSRs, they may agree to use label L to
represent Stream S for packets which are sent from Ru to Rd. That
is, they can agree to a "mapping" between label L and Stream S for
packets moving from Ru to Rd. As a result of such an agreement, L
becomes Ru's "outgoing label" corresponding to Stream S for such
packets; L becomes Rd's "incoming label" corresponding to Stream S
for such packets.
Note that L does not necessarily correspond to Stream S for any
packets other than those which are being sent from Ru to Rd. Also, L
is not an inherently meaningful value and does not have any network-
wide value; the particular value assigned to L gets its meaning
solely from the agreement between Ru and Rd.
Sometimes it may be difficult or even impossible for Rd to tell that
an arriving packet carrying label L comes from Ru, rather than from
some other LSR. In such cases, Rd must make sure that the mapping
from label to FEC is one-to-one. That is, in such cases, Rd must not
agree with Ru1 to use L for one purpose, while also agreeing with
some other LSR Ru2 to use L for a different purpose.
Rosen, Viswanathan & Callon [Page 10]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
The scope of labels could be unique per interface, or unique per MPLS
node, or unique in a network. If labels are unique within a network,
no label swapping needs to be performed in the MPLS nodes in that
domain. The packets are just label forwarded and not label swapped.
The possible use of labels with network-wide scope is FFS.
2.2. Upstream and Downstream LSRs
Suppose Ru and Rd have agreed to map label L to Stream S, for packets
sent from Ru to Rd. Then with respect to this mapping, Ru is the
"upstream LSR", and Rd is the "downstream LSR".
The notion of upstream and downstream relate to agreements between
nodes of the label values to be assigned for packets belonging to a
particular Stream that might be traveling from an upstream node to a
downstream node. This is independent of whether the routing protocol
actually will cause any packets to be transmitted in that particular
direction. Thus, Rd is the downstream LSR for a particular mapping
for label L if it recognizes L-labeled packets from Ru as being in
Stream S. This may be true even if routing does not actually forward
packets for Stream S between nodes Rd and Ru, or if routing has made
Ru downstream of Rd along the path which is actually used for packets
in Stream S.
2.3. Labeled Packet
A "labeled packet" is a packet into which a label has been encoded.
The encoding can be done by means of an encapsulation which exists
specifically for this purpose, or by placing the label in an
available location in either of the data link or network layer
headers. Of course, the encoding technique must be agreed to by the
entity which encodes the label and the entity which decodes the
label.
2.4. Label Assignment and Distribution; Attributes
For unicast traffic in the MPLS architecture, the decision to bind a
particular label L to a particular Stream S is made by the LSR which
is downstream with respect to that mapping. The downstream LSR then
informs the upstream LSR of the mapping. Thus labels are
"downstream-assigned", and are "distributed upstream".
A particular mapping of label L to Stream S, distributed by Rd to Ru,
may have associated "attributes". If Ru, acting as a downstream LSR,
also distributes a mapping of a label to Stream S, then under certain
Rosen, Viswanathan & Callon [Page 11]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
conditions, it may be required to also distribute the corresponding
attribute that it received from Rd.
2.5. Label Distribution Protocol (LDP)
A Label Distribution Protocol (LDP) is a set of procedures by which
one LSR informs another of the label/Stream mappings it has made.
Two LSRs which use an LDP to exchange label/Stream mapping
information are known as "LDP Peers" with respect to the mapping
information they exchange; we will speak of there being an "LDP
Adjacency" between them.
(N.B.: two LSRs may be LDP Peers with respect to some set of
mappings, but not with respect to some other set of mappings.)
The LDP also encompasses any negotiations in which two LDP Peers need
to engage in order to learn of each other's MPLS capabilities.
2.6. The Label Stack
So far, we have spoken as if a labeled packet carries only a single
label. As we shall see, it is useful to have a more general model in
which a labeled packet carries a number of labels, organized as a
last-in, first-out stack. We refer to this as a "label stack".
At a particular LSR, the decision as to how to forward a labeled
packet is always based exclusively on the label at the top of the
stack.
An unlabeled packet can be thought of as a packet whose label stack
is empty (i.e., whose label stack has depth 0).
If a packet's label stack is of depth m, we refer to the label at the
bottom of the stack as the level 1 label, to the label above it (if
such exists) as the level 2 label, and to the label at the top of the
stack as the level m label.
The utility of the label stack will become clear when we introduce
the notion of LSP Tunnel and the MPLS Hierarchy (sections 2.19.3 and
2.19.4).
Rosen, Viswanathan & Callon [Page 12]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2.7. The Next Hop Label Forwarding Entry (NHLFE)
The "Next Hop Label Forwarding Entry" (NHLFE) is used when forwarding
a labeled packet. It contains the following information:
1. the packet's next hop
2. the data link encapsulation to use when transmitting the packet
3. the way to encode the label stack when transmitting the packet
4. the operation to perform on the packet's label stack; this is
one of the following operations:
a) replace the label at the top of the label stack with a
specified new label
b) pop the label stack
c) replace the label at the top of the label stack with a
specified new label, and then push one or more specified
new labels onto the label stack.
Note that at a given LSR, the packet's "next hop" might be that LSR
itself. In this case, the LSR would need to pop the top level label
and examine and operate on the encapsulated packet. This may be a
lower level label, or may be the native IP packet. This implies that
in some cases the LSR may need to operate on the IP header in order
to forward the packet. If the packet's "next hop" is the current LSR,
then the label stack operation MUST be to "pop the stack".
2.8. Incoming Label Map (ILM)
The "Incoming Label Map" (ILM) is a mapping from incoming labels to
NHLFEs. It is used when forwarding packets that arrive as labeled
packets.
2.9. Stream-to-NHLFE Map (STN)
The "Stream-to-NHLFE" (STN) is a mapping from stream to NHLFEs. It is
used when forwarding packets that arrive unlabeled, but which are to
be labeled before being forwarded.
Rosen, Viswanathan & Callon [Page 13]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2.10. Label Swapping
Label swapping is the use of the following procedures to forward a
packet.
In order to forward a labeled packet, a LSR examines the label at the
top of the label stack. It uses the ILM to map this label to an
NHLFE. Using the information in the NHLFE, it determines where to
forward the packet, and performs an operation on the packet's label
stack. It then encodes the new label stack into the packet, and
forwards the result.
In order to forward an unlabeled packet, a LSR analyzes the network
layer header, to determine the packet's Stream. It then uses the FTN
to map this to an NHLFE. Using the information in the NHLFE, it
determines where to forward the packet, and performs an operation on
the packet's label stack. (Popping the label stack would, of course,
be illegal in this case.) It then encodes the new label stack into
the packet, and forwards the result.
It is important to note that when label swapping is in use, the next
hop is always taken from the NHLFE; this may in some cases be
different from what the next hop would be if MPLS were not in use.
2.11. Label Switched Path (LSP), LSP Ingress, LSP Egress
A "Label Switched Path (LSP) of level m" for a particular packet P is
a sequence of LSRs,
with the following properties:
1. R1, the "LSP Ingress", pushes a label onto P's label stack,
resulting in a label stack of depth m;
2. For all i, 1draft-ietf-mpls-arch-00.txt August 1997
5. For all i, 10).
In other words, we can speak of the level m LSP for Packet P as the
sequence of LSRs:
1. which begins with an LSR (an "LSP Ingress") that pushes on a
level m label,
2. all of whose intermediate LSRs make their forwarding decision
by label Switching on a level m label,
3. which ends (at an "LSP Egress") when a forwarding decision is
made by label Switching on a level m-k label, where k>0, or
when a forwarding decision is made by "ordinary", non-MPLS
forwarding procedures.
A consequence (or perhaps a presupposition) of this is that whenever
an LSR pushes a label onto an already labeled packet, it needs to
make sure that the new label corresponds to a FEC whose LSP Egress is
the LSR that assigned the label which is now second in the stack.
Note that according to these definitions, if is a level
m LSP for packet P, P may be transmitted from R[n-1] to Rn with a
label stack of depth m-1. That is, the label stack may be popped at
the penultimate LSR of the LSP, rather than at the LSP Egress. This
is appropriate, since the level m label has served its function of
getting the packet to Rn, and Rn's forwarding decision cannot be made
until the level m label is popped. If the label stack is not popped
by R[n-1], then Rn must do two label lookups; this is an overhead
which is best avoided. However, some hardware switching engines may
not be able to pop the label stack.
The penultimate node pops the label stack only if this is
specifically requested by the egress node. Having the penultimate
node pop the label stack has an implication on the assignment of
labels: For any one node Rn, operating at level m in the MPLS
Rosen, Viswanathan & Callon [Page 15]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
hierarchy, there may be some LSPs which terminate at that node (i.e.,
for which Rn is the egress node) and some other LSPs which continue
beyond that node (i.e., for which Rn is an intermediate node). If the
penultimate node R[n-1] pops the stack for those LSPs which terminate
at Rn, then node R[n] will receive some packets for which the top of
the stack is a level m label (i.e., packets destined for other egress
nodes), and some packets for which the top of the stack is a level
m-1 label (i.e., packets for which Rn is the egress). This implies
that in order for node R[n-1] to pop the stack, node Rn must assign
labels such that level m and level m-1 labels are distinguishable
(i.e., use unique values across multiple levels of the MPLS
hierarchy).
Note that if m = 1, the LSP Egress may receive an unlabeled packet,
and in fact need not even be capable of supporting MPLS. In this
case, assuming that we are using globally meaningful IP addresses,
the confusion of labels at multiple levels is not possible. However,
it is possible that the label may still be of value for the egress
node. One example is that the label may be used to assign the packet
to a particular Forwarding Equivalence Class (for example, to
identify the packet as a high priority packet). Another example is
that the label may assign the packet to a particular virtual private
network (for example, the virtual private network may make use of
local IP addresses, and the label may be necessary to disambiguate
the addresses). Therefore even when there is only a single label
value the stack is nonetheless popped only when requested by the
egress node.
We will call a sequence of LSRs the "LSP for a particular Stream S"
if it is an LSP of level m for a particular packet P when P's level m
label is a label corresponding to Stream S.
2.12. LSP Next Hop
The LSP Next Hop for a particular labeled packet in a particular LSR
is the LSR which is the next hop, as selected by the NHLFE entry used
for forwarding that packet.
The LSP Next Hop for a particular Stream is the next hop as selected
by the NHLFE entry indexed by a label which corresponds to that
Stream.
Rosen, Viswanathan & Callon [Page 16]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2.13. Route Selection
Route selection refers to the method used for selecting the LSP for a
particular stream. The proposed MPLS protocol architecture supports
two options for Route Selection: (1) Hop by hop routing, and (2)
Explicit routing.
Hop by hop routing allows each node to independently choose the next
hop for the path for a stream. This is the normal mode today with
existing datagram IP networks. A hop by hop routed LSP refers to an
LSP whose route is selected using hop by hop routing.
An explicitly routed LSP is an LSP where, at a given LSR, the LSP
next hop is not chosen by each local node, but rather is chosen by a
single node (usually the ingress or egress node of the LSP). The
sequence of LSRs followed by an explicit routing LSP may be chosen by
configuration, or by a protocol selected by a single node (for
example, the egress node may make use of the topological information
learned from a link state database in order to compute the entire
path for the tree ending at that egress node). Explicit routing may
be useful for a number of purposes such as allowing policy routing
and/or facilitating traffic engineering. With MPLS the explicit
route needs to be specified at the time that Labels are assigned, but
the explicit route does not have to be specified with each IP packet.
This implies that explicit routing with MPLS is relatively efficient
(when compared with the efficiency of explicit routing for pure
datagrams).
For any one LSP (at any one level of hierarchy), there are two
possible options: (i) The entire LSP may be hop by hop routed from
ingress to egress; (ii) The entire LSP may be explicit routed from
ingress to egress. Intermediate cases do not make sense: In general,
an LSP will be explicit routed specifically because there is a good
reason to use an alternative to the hop by hop routed path. This
implies that if some of the nodes along the path follow an explicit
route but some of the nodes make use of hop by hop routing, then
inconsistent routing will result and loops (or severely inefficient
paths) may form.
For this reason, it is important that if an explicit route is
specified for an LSP, then that route must be followed. Note that it
is relatively simple to *follow* an explicit route which is specified
in a LDP setup. We therefore propose that the LDP specification
require that all MPLS nodes implement the ability to follow an
explicit route if this is specified.
It is not necessary for a node to be able to create an explicit
route. However, in order to ensure interoperability it is necessary
Rosen, Viswanathan & Callon [Page 17]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
to ensure that either (i) Every node knows how to use hop by hop
routing; or (ii) Every node knows how to create and follow an
explicit route. We propose that due to the common use of hop by hop
routing in networks today, it is reasonable to make hop by hop
routing the default that all nodes need to be able to use.
2.14. Time-to-Live (TTL)
In conventional IP forwarding, each packet carries a "Time To Live"
(TTL) value in its header. Whenever a packet passes through a
router, its TTL gets decremented by 1; if the TTL reaches 0 before
the packet has reached its destination, the packet gets discarded.
This provides some level of protection against forwarding loops that
may exist due to misconfigurations, or due to failure or slow
convergence of the routing algorithm. TTL is sometimes used for other
functions as well, such as multicast scoping, and supporting the
"traceroute" command. This implies that there are two TTL-related
issues that MPLS needs to deal with: (i) TTL as a way to suppress
loops; (ii) TTL as a way to accomplish other functions, such as
limiting the scope of a packet.
When a packet travels along an LSP, it should emerge with the same
TTL value that it would have had if it had traversed the same
sequence of routers without having been label switched. If the
packet travels along a hierarchy of LSPs, the total number of LSR-
hops traversed should be reflected in its TTL value when it emerges
from the hierarchy of LSPs.
The way that TTL is handled may vary depending upon whether the MPLS
label values are carried in an MPLS-specific "shim" header, or if the
MPLS labels are carried in an L2 header such as an ATM header or a
frame relay header.
If the label values are encoded in a "shim" that sits between the
data link and network layer headers, then this shim should have a TTL
field that is initially loaded from the network layer header TTL
field, is decremented at each LSR-hop, and is copied into the network
layer header TTL field when the packet emerges from its LSP.
If the label values are encoded in an L2 header (e.g., the VPI/VCI
field in ATM's AAL5 header), and the labeled packets are forwarded by
an L2 switch (e.g., an ATM switch). This implies that unless the data
link layer itself has a TTL field (unlike ATM), it will not be
possible to decrement a packet's TTL at each LSR-hop. An LSP segment
which consists of a sequence of LSRs that cannot decrement a packet's
TTL will be called a "non-TTL LSP segment".
Rosen, Viswanathan & Callon [Page 18]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
When a packet emerges from a non-TTL LSP segment, it should however
be given a TTL that reflects the number of LSR-hops it traversed. In
the unicast case, this can be achieved by propagating a meaningful
LSP length to ingress nodes, enabling the ingress to decrement the
TTL value before forwarding packets into a non-TTL LSP segment.
Sometimes it can be determined, upon ingress to a non-TTL LSP
segment, that a particular packet's TTL will expire before the packet
reaches the egress of that non-TTL LSP segment. In this case, the LSR
at the ingress to the non-TTL LSP segment must not label switch the
packet. This means that special procedures must be developed to
support traceroute functionality, for example, traceroute packets may
be forwarded using conventional hop by hop forwarding.
2.15. Loop Control
On a non-TTL LSP segment, by definition, TTL cannot be used to
protect against forwarding loops. The importance of loop control may
depend on the particular hardware being used to provide the LSR
functions along the non-TTL LSP segment.
Suppose, for instance, that ATM switching hardware is being used to
provide MPLS switching functions, with the label being carried in the
VPI/VCI field. Since ATM switching hardware cannot decrement TTL,
there is no protection against loops. If the ATM hardware is capable
of providing fair access to the buffer pool for incoming cells
carrying different VPI/VCI values, this looping may not have any
deleterious effect on other traffic. If the ATM hardware cannot
provide fair buffer access of this sort, however, then even transient
loops may cause severe degradation of the LSR's total performance.
Even if fair buffer access can be provided, it is still worthwhile to
have some means of detecting loops that last "longer than possible".
In addition, even where TTL and/or per-VC fair queuing provides a
means for surviving loops, it still may be desirable where practical
to avoid setting up LSPs which loop.
The MPLS architecture will therefore provide a technique for ensuring
that looping LSP segments can be detected, and a technique for
ensuring that looping LSP segments are never created.
Rosen, Viswanathan & Callon [Page 19]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2.15.1. Loop Prevention
LSR's maintain for each of their LSP's an LSR id list. This list is a
list of all the LSR's downstream from this LSR on a given LSP. The
LSR id list is used to prevent the formation of switched path loops.
The LSR ID list is propagated upstream from a node to its neighbor
nodes. The LSR ID list is used to prevent loops as follows:
When a node, R, detects a change in the next hop for a given stream,
it asks its new next hop for a label and the associated LSR ID list
for that stream.
The new next hop responds with a label for the stream and an
associated LSR id list.
R looks in the LSR id list. If R determines that it, R, is in the
list then we have a route loop. In this case, we do nothing and the
old LSP will continue to be used until the route protocols break the
loop. The means by which the old LSP is replaced by a new LSP after
the route protocols breathe loop is described below.
If R is not in the LSR id list, R will start a "diffusion"
computation [12]. The purpose of the diffusion computation is to
prune the tree upstream of R so that we remove all LSR's from the
tree that would be on a looping path if R were to switch over to the
new LSP. After those LSR's are removed from the tree, it is safe for
R to replace the old LSP with the new LSP (and the old LSP can be
released).
The diffusion computation works as follows:
R adds its LSR id to the list and sends a query message to each of
its "upstream" neighbors (i.e. to each of its neighbors that is not
the new "downstream" next hop).
A node S that receives such a query will process the query as
follows:
- If node R is not node S's next hop for the given stream, node S
will respond to node R will an "OK" message meaning that as far
as node S is concerned it is safe for node R to switch over to
the new LSP.
- If node R is node S's next hop for the stream, node S will check
to see if it, node S, is in the LSR id list that it received from
node R. If it is, we have a route loop and S will respond with a
"LOOP" message. R will unsplice the connection to S pruning S
from the tree. The mechanism by which S will get a new LSP for
Rosen, Viswanathan & Callon [Page 20]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
the stream after the route protocols break the loop is described
below.
- If node S is not in the LSR id list, S will add its LSR id to the
LSR id list and send a new query message further upstream. The
diffusion computation will continue to propagate upstream along
each of the paths in the tree upstream of S until either a loop
is detected, in which case the node is pruned as described above
or we get to a point where a node gets a response ("OK" or
"LOOP") from each of its neighbors perhaps because none of those
neighbors considers the node in question to be its downstream
next hop. Once a node has received a response from each of its
upstream neighbors, it returns an "OK" message to its downstream
neighbor. When the original node, node R, gets a response from
each of its neighbors, it is safe to replace the old LSP with the
new one because all the paths that would loop have been pruned
from the tree.
There are a couple of details to discuss:
- First, we need to do something about nodes that for one reason or
another do not produce a timely response in response to a query
message. If a node Y does not respond to a query from node X
because of a failure of some kind, X will not be able to respond
to its downstream neighbors (if any) or switch over to a new LSP
if X is, like R above, the node that has detected the route
change. This problem is handled by timing out the query message.
If a node doesn't receive a response within a "reasonable" period
of time, it "unsplices" its VC to the upstream neighbor that is
not responding and proceeds as it would if it had received the
"LOOP" message.
- We also need to be concerned about multiple concurrent routing
updates. What happens, for example, when a node M receives a
request for an LSP from an upstream neighbor, N, when M is in the
middle of a diffusion computation i.e., it has sent a query
upstream but hasn't received all the responses. Since a
downstream node, node R is about to change from one LSP to
another, M needs to pass to N an LSR id list corresponding to the
union of the old and new LSP's if it is to avoid loops both
before and after the transition. This is easily accomplished
since M already has the LSR id list for the old LSP and it gets
the LSR id list for the new LSP in the query message. After R
makes the switch from the old LSP to the new one, R sends a new
establish message upstream with the LSR id list of (just) the new
LSP. At this point, the nodes upstream of R know that R has
switched over to the new LSP and that they can return the id list
for (just) the new LSP in response to any new requests for LSP's.
Rosen, Viswanathan & Callon [Page 21]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
They can also grow the tree to include additional nodes that
would not have been valid for the combined LSR id list.
- We also need to discuss how a node that doesn't have an LSP for a
given stream at the end of a diffusion computation (because it
would have been on a looping LSP) gets one after the routing
protocols break the loop. If node L has been pruned from the
tree and its local route protocol processing entity breaks the
loop by changing L's next hop, L will request a new LSP from its
new downstream neighbor which it will use once it executes the
diffusion computation as described above. If the loop is broken
by a route change at another point in the loop, i.e. at a point
"downstream" of L, L will get a new LSP as the new LSP tree grows
upstream from the point of the route change as discussed in the
previous paragraph.
- Note that when a node is pruned from the tree, the switched path
upstream of that node remains "connected". This is important
since it allows the switched path to get "reconnected" to a
downstream switched path after a route change with a minimal
amount of unsplicing and resplicing once the appropriate
diffusion computation(s) have taken place.
The LSR Id list can also be used to provide a "loop detection"
capability. To use it in this manner, an LSR which sees that it is
already in the LSR Id list for a particular stream will immediately
unsplice itself from the switched path for that stream, and will NOT
pass the LSR Id list further upstream. The LSR can rejoin a switched
path for the stream when it changes its next hop for that stream, or
when it receives a new LSR Id list from its current next hop, in
which it is not contained. The diffusion computation would be
omitted.
2.15.2. Interworking of Loop Control Options
The MPLS protocol architecture allows some nodes to be using loop
prevention, while some other nodes are not (i.e., the choice of
whether or not to use loop prevention may be a local decision). When
this mix is used, it is not possible for a loop to form which
includes only nodes which do loop prevention. However, it is possible
for loops to form which contain a combination of some nodes which do
loop prevention, and some nodes which do not.
There are at least four identified cases in which it makes sense to
combine nodes which do loop prevention with nodes which do not: (i)
For transition, in intermediate states while transitioning from all
non-loop-prevention to all loop prevention, or vice versa; (ii) For
Rosen, Viswanathan & Callon [Page 22]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
interoperability, where one vendor implements loop prevention but
another vendor does not; (iii) Where there is a mixed ATM and
datagram media network, and where loop prevention is desired over the
ATM portions of the network but not over the datagram portions; (iv)
where some of the ATM switches can do fair access to the buffer pool
on a per-VC basis, and some cannot, and loop prevention is desired
over the ATM portions of the network which cannot.
Note that interworking is straightforward. If an LSR is not doing
loop prevention, and it receives from a downstream LSR a label
mapping which contains loop prevention information, it (a) accepts
the label mapping, (b) does NOT pass the loop prevention information
upstream, and (c) informs the downstream neighbor that the path is
loop-free.
Similarly, if an LSR R which is doing loop prevention receives from a
downstream LSR a label mapping which does not contain any loop
prevention information, then R passes the label mapping upstream with
loop prevention information included as if R were the egress for the
specified stream.
Optionally, a node is permitted to implement the ability of either
doing or not doing loop prevention as options, and is permitted to
choose which to use for any one particular LSP based on the
information obtained from downstream nodes. When the label mapping
arrives from downstream, then the node may choose whether to use loop
prevention so as to continue to use the same approach as was used in
the information passed to it. Note that regardless of whether loop
prevention is used the egress nodes (for any particular LSP) always
initiates exchange of label mapping information without waiting for
other nodes to act.
2.16. Merging and Non-Merging LSRs
Merge allows multiple upstream LSPs to be merged into a single
downstream LSP. When implemented by multiple nodes, this results in
the traffic going to a particular egress nodes, based on one
particular Stream, to follow a multipoint to point tree (MPT), with
the MPT rooted at the egress node and associated with the Stream.
This can have a significant effect on reducing the number of labels
that need to be maintained by any one particular node.
If merge was not used at all it would be necessary for each node to
provide the upstream neighbors with a label for each Stream for each
upstream node which may be forwarding traffic over the link. This
implies that the number of labels needed might not in general be
known a priori. However, the use of merge allows a single label to be
Rosen, Viswanathan & Callon [Page 23]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
used per Stream, therefore allowing label assignment to be done in a
common way without regard for the number of upstream nodes which will
be using the downstream LSP.
The proposed MPLS protocol architecture supports LSP merge, while
allowing nodes which do not support LSP merge. This leads to the
issue of ensuring correct interoperation between nodes which
implement merge and those which do not. The issue is somewhat
different in the case of datagram media versus the case of ATM. The
different media types will therefore be discussed separately.
2.16.1. Stream Merge
Let us say that an LSR is capable of Stream Merge if it can receive
two packets from different incoming interfaces, and/or with different
labels, and send both packets out the same outgoing interface with
the same label. This in effect takes two incoming streams and merges
them into one. Once the packets are transmitted, the information that
they arrived from different interfaces and/or with different incoming
labels is lost.
Let us say that an LSR is not capable of Stream Merge if, for any two
packets which arrive from different interfaces, or with different
labels, the packets must either be transmitted out different
interfaces, or must have different labels.
An LSR which is capable of Stream Merge (a "Merging LSR") needs to
maintain only one outgoing label for each FEC. AN LSR which is not
capable of Stream Merge (a "Non-merging LSR") may need to maintain as
many as N outgoing labels per FEC, where N is the number of LSRs in
the network. Hence by supporting Stream Merge, an LSR can reduce its
number of outgoing labels by a factor of O(N). Since each label in
use requires the dedication of some amount of resources, this can be
a significant savings.
2.16.2. Non-merging LSRs
The MPLS forwarding procedures is very similar to the forwarding
procedures used by such technologies as ATM and Frame Relay. That is,
a unit of data arrives, a label (VPI/VCI or DLCI) is looked up in a
"cross-connect table", on the basis of that lookup an output port is
chosen, and the label value is rewritten. In fact, it is possible to
use such technologies for MPLS forwarding; LDP can be used as the
"signalling protocol" for setting up the cross-connect tables.
Unfortunately, these technologies do not necessarily support the
Rosen, Viswanathan & Callon [Page 24]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
Stream Merge capability. In ATM, if one attempts to perform Stream
Merge, the result may be the interleaving of cells from various
packets. If cells from different packets get interleaved, it is
impossible to reassemble the packets. Some Frame Relay switches use
cell switching on their backplanes. These switches may also be
incapable of supporting Stream Merge, for the same reason -- cells of
different packets may get interleaved, and there is then no way to
reassemble the packets.
We propose to support two solutions to this problem. First, MPLS will
contain procedures which allow the use of non-merging LSRs. Second,
MPLS will support procedures which allow certain ATM switches to
function as merging LSRs.
Since MPLS supports both merging and non-merging LSRs, MPLS also
contains procedures to ensure correct interoperation between them.
2.16.3. Labels for Merging and Non-Merging LSRs
An upstream LSR which supports Stream Merge needs to be sent only one
label per FEC. An upstream neighbor which does not support Stream
Merge needs to be sent multiple labels per FEC. However, there is no
way of knowing a priori how many labels it needs. This will depend on
how many LSRs are upstream of it with respect to the FEC in question.
In the MPLS architecture, if a particular upstream neighbor does not
support Stream Merge, it is not sent any labels for a particular FEC
unless it explicitly asks for a label for that FEC. The upstream
neighbor may make multiple such requests, and is given a new label
each time. When a downstream neighbor receives such a request from
upstream, and the downstream neighbor does not itself support Stream
Merge, then it must in turn ask its downstream neighbor for another
label for the FEC in question.
It is possible that there may be some nodes which support merge, but
have a limited number of upstream streams which may be merged into a
single downstream streams. Suppose for example that due to some
hardware limitation a node is capable of merging four upstream LSPs
into a single downstream LSP. Suppose however, that this particular
node has six upstream LSPs arriving at it for a particular Stream. In
this case, this node may merge these into two downstream LSPs
(corresponding to two labels that need to be obtained from the
downstream neighbor). In this case, the normal operation of the LDP
implies that the downstream neighbor will supply this node with a
single label for the Stream. This node can then ask its downstream
neighbor for one additional label for the Stream, implying that the
node will thereby obtain the required two labels.
Rosen, Viswanathan & Callon [Page 25]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
The interaction between explicit routing and merge is FFS.
2.16.4. Merge over ATM
2.16.4.1. Methods of Eliminating Cell Interleave
There are several methods that can be used to eliminate the cell
interleaving problem in ATM, thereby allowing ATM switches to support
stream merge: :
1. VP merge
When VP merge is used, multiple virtual paths are merged into a
virtual path, but packets from different sources are
distinguished by using different VCs within the VP.
2. VC merge
When VC merge is used, switches are required to buffer cells
from one packet until the entire packet is received (this may
be determined by looking for the AAL5 end of frame indicator).
VP merge has the advantage that it is compatible with a higher
percentage of existing ATM switch implementations. This makes it more
likely that VP merge can be used in existing networks. Unlike VC
merge, VP merge does not incur any delays at the merge points and
also does not impose any buffer requirements. However, it has the
disadvantage that it requires coordination of the VCI space within
each VP. There are a number of ways that this can be accomplished.
Selection of one or more methods is FFS.
This tradeoff between compatibility with existing equipment versus
protocol complexity and scalability implies that it is desirable for
the MPLS protocol to support both VP merge and VC merge. In order to
do so each ATM switch participating in MPLS needs to know whether its
immediate ATM neighbors perform VP merge, VC merge, or no merge.
2.16.4.2. Interoperation: VC Merge, VP Merge, and Non-Merge
The interoperation of the various forms of merging over ATM is most
easily described by first describing the interoperation of VC merge
with non-merge.
In the case where VC merge and non-merge nodes are interconnected the
forwarding of cells is based in all cases on a VC (i.e., the
concatenation of the VPI and VCI). For each node, if an upstream
Rosen, Viswanathan & Callon [Page 26]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
neighbor is doing VC merge then that upstream neighbor requires only
a single VPI/VCI for a particular Stream (this is analogous to the
requirement for a single label in the case of operation over frame
media). If the upstream neighbor is not doing merge, then the
neighbor will require a single VPI/VCI per Stream for itself, plus
enough VPI/VCIs to pass to its upstream neighbors. The number
required will be determined by allowing the upstream nodes to request
additional VPI/VCIs from their downstream neighbors (this is again
analogous to the method used with frame merge).
A similar method is possible to support nodes which perform VP merge.
In this case the VP merge node, rather than requesting a single
VPI/VCI or a number of VPI/VCIs from its downstream neighbor, instead
may request a single VP (identified by a VPI) but several VCIs within
the VP. Furthermore, suppose that a non-merge node is downstream
from two different VP merge nodes. This node may need to request one
VPI/VCI (for traffic originating from itself) plus two VPs (one for
each upstream node), each associated with a specified set of VCIs (as
requested from the upstream node).
In order to support all of VP merge, VC merge, and non-merge, it is
therefore necessary to allow upstream nodes to request a combination
of zero or more VC identifiers (consisting of a VPI/VCI), plus zero
or more VPs (identified by VPIs) each containing a specified number
of VCs (identified by a set of VCIs which are significant within a
VP). VP merge nodes would therefore request one VP, with a contained
VCI for traffic that it originates (if appropriate) plus a VCI for
each VC requested from above (regardless of whether or not the VC is
part of a containing VP). VC merge node would request only a single
VPI/VCI (since they can merge all upstream traffic into a single VC).
Non-merge nodes would pass on any requests that they get from above,
plus request a VPI/VCI for traffic that they originate (if
appropriate).
2.17. LSP Control: Egress versus Local
There is a choice to be made regarding whether the initial setup of
LSPs will be initiated by the egress node, or locally by each
individual node.
When LSP control is done locally, then each node may at any time pass
label bindings to its neighbors for each FEC recognized by that node.
In the normal case that the neighboring nodes recognize the same
FECs, then nodes may map incoming labels to outgoing labels as part
of the normal label swapping forwarding method.
When LSP control is done by the egress, then initially only the
Rosen, Viswanathan & Callon [Page 27]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
egress node passes label bindings to its neighbors corresponding to
any FECs which leave the MPLS network at that egress node. Other
nodes wait until they get a label from downstream for a particular
FEC before passing a corresponding label for the same FEC to upstream
nodes.
With local control, since each LSR is (at least initially)
independently assigning labels to FECs, it is possible that different
LSRs may make inconsistent decisions. For example, an upstream LSR
may make a coarse decision (map multiple IP address prefixes to a
single label) while its downstream neighbor makes a finer grain
decision (map each individual IP address prefix to a separate label).
With downstream label assignment this can be corrected by having LSRs
withdraw labels that it has assigned which are inconsistent with
downstream labels, and replace them with new consistent label
assignments.
Even with egress control it is possible that the choice of egress
node may change, or the egress may (based on a change in
configuration) change its mind in terms of the granularity which is
to be used. This implies the same mechanism will be necessary to
allow changes in granularity to bubble up to upstream nodes. The
choice of egress or local control may therefore effect the frequency
with which this mechanism is used, but will not effect the need for a
mechanism to achieve consistency of label granularity. Generally
speaking, the choice of local versus egress control does not appear
to have any effect on the LDP mechanisms which need to be defined.
Egress control and local control can interwork in a very
straightforward manner (although some of the advantages ascribed to
egress control may be lost, see appendices A and B). With either
approach, (assuming downstream label assignment) the egress node will
initially assign labels for particular FECs and will pass these
labels to its neighbors. With either approach these label assignments
will bubble upstream, with the upstream nodes choosing labels that
are consistent with the labels that they receive from downstream. The
difference between the two approaches is therefore primarily an issue
of what each node does prior to obtaining a label assignment for a
particular FEC from downstream nodes: Does it wait, or does it assign
a preliminary label under the expectation that it will (probably) be
correct?
Regardless of which method is used (local control or egress control)
each node needs to know (possibly by configuration) what granularity
to use for labels that it assigns. Where egress control is used, this
requires each node to know the granularity only for streams which
leave the MPLS network at that node. For local control, in order to
avoid the need to withdraw inconsistent labels, each node in the
Rosen, Viswanathan & Callon [Page 28]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
network would need to be configured consistently to know the
granularity for each stream. However, in many cases this may be done
by using a single level of granularity which applies to all streams
(such as "one label per IP prefix in the forwarding table"). The
choice between local control versus egress control could similarly be
left as a configuration option.
Future versions of the MPLS architecture will need to choose between
three options: (i) Requiring local control; (ii) Requiring egress
control; or (iii) Allowing a choice of local control or egress
control. Arguments for local versus egress control are contained in
appendices A and B.
2.18. Granularity
When forwarding by label swapping, a stream of packets following a
stream arriving from upstream may be mapped into an equal or coarser
grain stream. However, a coarse grain stream (for example, containing
packets destined for a short IP address prefix covering many subnets)
cannot be mapped directly into a finer grain stream (for example,
containing packets destined for a longer IP address prefix covering a
single subnet). This implies that there needs to be some mechanism
for ensuring consistency between the granularity of LSPs in an MPLS
network.
The method used for ensuring compatibility of granularity may depend
upon the method used for LSP control.
When LSP control is local, it is possible that a node may pass a
coarse grain label to its upstream neighbor(s), and subsequently
receive a finer grain label from its downstream neighbor. In this
case the node has two options: (i) It may forward the corresponding
packets using normal IP datagram forwarding (i.e., by examination of
the IP header); (ii) It may withdraw the label mappings that it has
passed to its upstream neighbors, and replace these with finer grain
label mappings.
When LSP control is egress based, the label setup originates from the
egress node and passes upstream. It is therefore straightforward with
this approach to maintain equally-grained mappings along the route.
Rosen, Viswanathan & Callon [Page 29]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2.19. Tunnels and Hierarchy
Sometimes a router Ru takes explicit action to cause a particular
packet to be delivered to another router Rd, even though Ru and Rd
are not consecutive routers on the Hop-by-hop path for that packet,
and Rd is not the packet's ultimate destination. For example, this
may be done by encapsulating the packet inside a network layer packet
whose destination address is the address of Rd itself. This creates a
"tunnel" from Ru to Rd. We refer to any packet so handled as a
"Tunneled Packet".
2.19.1. Hop-by-Hop Routed Tunnel
If a Tunneled Packet follows the Hop-by-hop path from Ru to Rd, we
say that it is in an "Hop-by-Hop Routed Tunnel" whose "transmit
endpoint" is Ru and whose "receive endpoint" is Rd.
2.19.2. Explicitly Routed Tunnel
If a Tunneled Packet travels from Ru to Rd over a path other than the
Hop-by-hop path, we say that it is in an "Explicitly Routed Tunnel"
whose "transmit endpoint" is Ru and whose "receive endpoint" is Rd.
For example, we might send a packet through an Explicitly Routed
Tunnel by encapsulating it in a packet which is source routed.
2.19.3. LSP Tunnels
It is possible to implement a tunnel as a LSP, and use label
switching rather than network layer encapsulation to cause the packet
to travel through the tunnel. The tunnel would be a LSP , where R1 is the transmit endpoint of the tunnel, and Rn is the
receive endpoint of the tunnel. This is called a "LSP Tunnel".
The set of packets which are to be sent though the LSP tunnel becomes
a Stream, and each LSR in the tunnel must assign a label to that
Stream (i.e., must assign a label to the tunnel). The criteria for
assigning a particular packet to an LSP tunnel is a local matter at
the tunnel's transmit endpoint. To put a packet into an LSP tunnel,
the transmit endpoint pushes a label for the tunnel onto the label
stack and sends the labeled packet to the next hop in the tunnel.
If it is not necessary for the tunnel's receive endpoint to be able
to determine which packets it receives through the tunnel, as
discussed earlier, the label stack may be popped at the penultimate
LSR in the tunnel.
Rosen, Viswanathan & Callon [Page 30]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
A "Hop-by-Hop Routed LSP Tunnel" is a Tunnel that is implemented as
an hop-by-hop routed LSP between the transmit endpoint and the
receive endpoint.
An "Explicitly Routed LSP Tunnel" is a LSP Tunnel that is also an
Explicitly Routed LSP.
2.19.4. Hierarchy: LSP Tunnels within LSPs
Consider a LSP . Let us suppose that R1 receives
unlabeled packet P, and pushes on its label stack the label to cause
it to follow this path, and that this is in fact the Hop-by-hop path.
However, let us further suppose that R2 and R3 are not directly
connected, but are "neighbors" by virtue of being the endpoints of an
LSP tunnel. So the actual sequence of LSRs traversed by P is .
When P travels from R1 to R2, it will have a label stack of depth 1.
R2, switching on the label, determines that P must enter the tunnel.
R2 first replaces the Incoming label with a label that is meaningful
to R3. Then it pushes on a new label. This level 2 label has a value
which is meaningful to R21. Switching is done on the level 2 label by
R21, R22, R23. R23, which is the penultimate hop in the R2-R3 tunnel,
pops the label stack before forwarding the packet to R3. When R3 sees
packet P, P has only a level 1 label, having now exited the tunnel.
Since R3 is the penultimate hop in P's level 1 LSP, it pops the label
stack, and R4 receives P unlabeled.
The label stack mechanism allows LSP tunneling to nest to any depth.
2.19.5. LDP Peering and Hierarchy
Suppose that packet P travels along a Level 1 LSP ,
and when going from R2 to R3 travels along a Level 2 LSP . From the perspective of the Level 2 LSP, R2's LDP peer is
R21. From the perspective of the Level 1 LSP, R2's LDP peers are R1
and R3. One can have LDP peers at each layer of hierarchy. We will
see in sections 3.6 and 3.7 some ways to make use of this hierarchy.
Note that in this example, R2 and R21 must be IGP neighbors, but R2
and R3 need not be.
When two LSRs are IGP neighbors, we will refer to them as "Local LDP
Peers". When two LSRs may be LDP peers, but are not IGP neighbors,
we will refer to them as "Remote LDP Peers". In the above example,
R2 and R21 are local LDP peers, but R2 and R3 are remote LDP peers.
Rosen, Viswanathan & Callon [Page 31]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
The MPLS architecture supports two ways to distribute labels at
different layers of the hierarchy: Explicit Peering and Implicit
Peering.
One performs label Distribution with one's Local LDP Peers by opening
LDP connections to them. One can perform label Distribution with
one's Remote LDP Peers in one of two ways:
1. Explicit Peering
In explicit peering, one sets up LDP connections between Remote
LDP Peers, exactly as one would do for Local LDP Peers. This
technique is most useful when the number of Remote LDP Peers is
small, or the number of higher level label mappings is large,
or the Remote LDP Peers are in distinct routing areas or
domains. Of course, one needs to know which labels to
distribute to which peers; this is addressed in section 3.1.2.
Examples of the use of explicit peering is found in sections
3.2.1 and 3.6.
2. Implicit Peering
In Implicit Peering, one does not have LDP connections to one's
remote LDP peers, but only to one's local LDP peers. To
distribute higher level labels to ones remote LDP peers, one
encodes the higher level labels as an attribute of the lower
level labels, and distributes the lower level label, along with
this attribute, to the local LDP peers. The local LDP peers
then propagate the information to their peers. This process
continues till the information reaches remote LDP peers. Note
that the intermediary nodes may also be remote LDP peers.
This technique is most useful when the number of Remote LDP
Peers is large. Implicit peering does not require a n-square
peering mesh to distribute labels to the remote LDP peers
because the information is piggybacked through the local LDP
peering. However, implicit peering requires the intermediate
nodes to store information that they might not be directly
interested in.
An example of the use of implicit peering is found in section
3.3.
Rosen, Viswanathan & Callon [Page 32]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2.20. LDP Transport
LDP is used between nodes in an MPLS network to establish and
maintain the label mappings. In order for LDP to operate correctly,
LDP information needs to be transmitted reliably, and the LDP
messages pertaining to a particular FEC need to be transmitted in
sequence. This may potentially be accomplished either by using an
existing reliable transport protocol such as TCP, or by specifying
reliability mechanisms as part of LDP (for example, the reliability
mechanisms which are defined in IDRP could potentially be "borrowed"
for use with LSP). The precise means for accomplishing transport
reliability with LSP are for further study, but will be specified by
the MPLS Protocol Architecture before the architecture may be
considered complete.
2.21. Label Encodings
In order to transmit a label stack along with the packet whose label
stack it is, it is necessary to define a concrete encoding of the
label stack. The architecture supports several different encoding
techniques; the choice of encoding technique depends on the
particular kind of device being used to forward labeled packets.
2.21.1. MPLS-specific Hardware and/or Software
If one is using MPLS-specific hardware and/or software to forward
labeled packets, the most obvious way to encode the label stack is to
define a new protocol to be used as a "shim" between the data link
layer and network layer headers. This shim would really be just an
encapsulation of the network layer packet; it would be "protocol-
independent" such that it could be used to encapsulate any network
layer. Hence we will refer to it as the "generic MPLS
encapsulation".
The generic MPLS encapsulation would in turn be encapsulated in a
data link layer protocol.
The generic MPLS encapsulation should contain the following fields:
1. the label stack,
2. a Time-to-Live (TTL) field
Rosen, Viswanathan & Callon [Page 33]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
3. a Class of Service (CoS) field
The TTL field permits MPLS to provide a TTL function similar to what
is provided by IP.
The CoS field permits LSRs to apply various scheduling packet
disciplines to labeled packets, without requiring separate labels for
separate disciplines.
This section is not intended to rule out the use of alternative
mechanisms in network environments where such alternatives may be
appropriate.
2.21.2. ATM Switches as LSRs
It will be noted that MPLS forwarding procedures are similar to those
of legacy "label swapping" switches such as ATM switches. ATM
switches use the input port and the incoming VPI/VCI value as the
index into a "cross-connect" table, from which they obtain an output
port and an outgoing VPI/VCI value. Therefore if one or more labels
can be encoded directly into the fields which are accessed by these
legacy switches, then the legacy switches can, with suitable software
upgrades, be used as LSRs. We will refer to such devices as "ATM-
LSRs".
There are three obvious ways to encode labels in the ATM cell header
(presuming the use of AAL5):
1. SVC Encoding
Use the VPI/VCI field to encode the label which is at the top
of the label stack. This technique can be used in any network.
With this encoding technique, each LSP is realized as an ATM
SVC, and the LDP becomes the ATM "signaling" protocol. With
this encoding technique, the ATM-LSRs cannot perform "push" or
"pop" operations on the label stack.
2. SVP Encoding
Use the VPI field to encode the label which is at the top of
the label stack, and the VCI field to encode the second label
on the stack, if one is present. This technique some advantages
over the previous one, in that it permits the use of ATM "VP-
switching". That is, the LSPs are realized as ATM SVPs, with
LDP serving as the ATM signaling protocol.
However, this technique cannot always be used. If the network
Rosen, Viswanathan & Callon [Page 34]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
includes an ATM Virtual Path through a non-MPLS ATM network,
then the VPI field is not necessarily available for use by
MPLS.
When this encoding technique is used, the ATM-LSR at the egress
of the VP effectively does a "pop" operation.
3. SVP Multipoint Encoding
Use the VPI field to encode the label which is at the top of
the label stack, use part of the VCI field to encode the second
label on the stack, if one is present, and use the remainder of
the VCI field to identify the LSP ingress. If this technique
is used, conventional ATM VP-switching capabilities can be used
to provide multipoint-to-point VPs. Cells from different
packets will then carry different VCI values, so multipoint-
to-point VPs can be provided without any cell interleaving
problems.
This technique depends on the existence of a capability for
assigning small unique values to each ATM switch.
If there are more labels on the stack than can be encoded in the ATM
header, the ATM encodings must be combined with the generic
encapsulation. This does presuppose that it be possible to tell,
when reassembling the ATM cells into packets, whether the generic
encapsulation is also present.
2.21.3. Interoperability among Encoding Techniques
If is a segment of a LSP, it is possible that R1 will
use one encoding of the label stack when transmitting packet P to R2,
but R2 will use a different encoding when transmitting a packet P to
R3. In general, the MPLS architecture supports LSPs with different
label stack encodings used on different hops. Therefore, when we
discuss the procedures for processing a labeled packet, we speak in
abstract terms of operating on the packet's label stack. When a
labeled packet is received, the LSR must decode it to determine the
current value of the label stack, then must operate on the label
stack to determine the new value of the stack, and then encode the
new value appropriately before transmitting the labeled packet to its
next hop.
Unfortunately, ATM switches have no capability for translating from
one encoding technique to another. The MPLS architecture therefore
requires that whenever it is possible for two ATM switches to be
successive LSRs along a level m LSP for some packet, that those two
Rosen, Viswanathan & Callon [Page 35]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
ATM switches use the same encoding technique.
Naturally there will be MPLS networks which contain a combination of
ATM switches operating as LSRs, and other LSRs which operate using an
MPLS shim header. In such networks there may be some LSRs which have
ATM interfaces as well as "MPLS Shim" interfaces. This is one example
of an LSR with different label stack encodings on different hops.
Such an LSR may swap off an ATM encoded label stack on an incoming
interface and replace it with an MPLS shim header encoded label stack
on the outgoing interface.
2.22. Multicast
This section is for further study
3. Some Applications of MPLS
3.1. MPLS and Hop by Hop Routed Traffic
One use of MPLS is to simplify the process of forwarding packets
using hop by hop routing.
3.1.1. Labels for Address Prefixes
In general, router R determines the next hop for packet P by finding
the address prefix X in its routing table which is the longest match
for P's destination address. That is, the packets in a given Stream
are just those packets which match a given address prefix in R's
routing table. In this case, a Stream can be identified with an
address prefix.
If packet P must traverse a sequence of routers, and at each router
in the sequence P matches the same address prefix, MPLS simplifies
the forwarding process by enabling all routers but the first to avoid
executing the best match algorithm; they need only look up the label.
3.1.2. Distributing Labels for Address Prefixes
3.1.2.1. LDP Peers for a Particular Address Prefix
LSRs R1 and R2 are considered to be LDP Peers for address prefix X if
and only if one of the following conditions holds:
Rosen, Viswanathan & Callon [Page 36]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
1. R1's route to X is a route which it learned about via a
particular instance of a particular IGP, and R2 is a neighbor
of R1 in that instance of that IGP
2. R1's route to X is a route which it learned about by some
instance of routing algorithm A1, and that route is
redistributed into an instance of routing algorithm A2, and R2
is a neighbor of R1 in that instance of A2
3. R1 is the receive endpoint of an LSP Tunnel that is within
another LSP, and R2 is a transmit endpoint of that tunnel, and
R1 and R2 are participants in a common instance of an IGP, and
are in the same IGP area (if the IGP in question has areas),
and R1's route to X was learned via that IGP instance, or is
redistributed by R1 into that IGP instance
4. R1's route to X is a route which it learned about via BGP, and
R2 is a BGP peer of R1
In general, these rules ensure that if the route to a particular
address prefix is distributed via an IGP, the LDP peers for that
address prefix are the IGP neighbors. If the route to a particular
address prefix is distributed via BGP, the LDP peers for that address
prefix are the BGP peers. In other cases of LSP tunneling, the
tunnel endpoints are LDP peers.
3.1.2.2. Distributing Labels
In order to use MPLS for the forwarding of normally routed traffic,
each LSR MUST:
1. bind one or more labels to each address prefix that appears in
its routing table;
2. for each such address prefix X, use an LDP to distribute the
mapping of a label to X to each of its LDP Peers for X.
There is also one circumstance in which an LSR must distribute a
label mapping for an address prefix, even if it is not the LSR which
bound that label to that address prefix:
3. If R1 uses BGP to distribute a route to X, naming some other
LSR R2 as the BGP Next Hop to X, and if R1 knows that R2 has
assigned label L to X, then R1 must distribute the mapping
between T and X to any BGP peer to which it distributes that
route.
Rosen, Viswanathan & Callon [Page 37]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
These rules ensure that labels corresponding to address prefixes
which correspond to BGP routes are distributed to IGP neighbors if
and only if the BGP routes are distributed into the IGP. Otherwise,
the labels bound to BGP routes are distributed only to the other BGP
speakers.
These rules are intended to indicate which label mappings must be
distributed by a given LSR to which other LSRs, NOT to indicate the
conditions under which the distribution is to be made. That is
discussed in section 2.17.
3.1.3. Using the Hop by Hop path as the LSP
If the hop-by-hop path that packet P needs to follow is , then can be an LSP as long as:
1. there is a single address prefix X, such that, for all i,
1<=idraft-ietf-mpls-arch-00.txt August 1997
2. R contains in its routing tables one or more address prefixes Y
such that X is a proper initial substring of Y, but R's "LSP
previous hops" for X do not contain any such address prefixes
Y; that is, R2 is a "deaggregation point" for address prefix X.
An LSR R1 is considered to be an "LSP Proxy Egress" LSR for address
prefix X if and only if:
1. R1's next hop for X is R2 R1 and R2 are not LDP Peers with
respect to X (perhaps because R2 does not support MPLS), or
2. R1 has been configured to act as an LSP Proxy Egress for X
The definition of LSP allows for the LSP Egress to be a node which
does not support MPLS; in this case the penultimate node in the LSP
is the Proxy Egress.
3.1.5. The POP Label
The POP label is a label with special semantics which an LSR can bind
to an address prefix. If LSR Ru, by consulting its ILM, sees that
labeled packet P must be forwarded next to Rd, but that Rd has
distributed a mapping of the POP label to the corresponding address
prefix, then instead of replacing the value of the label on top of
the label stack, Ru pops the label stack, and then forwards the
resulting packet to Rd.
LSR Rd distributes a mapping between the POP label and an address
prefix X to LSR Ru if and only if:
1. the rules of Section 3.1.2 indicate that Rd distributes to Ru a
label mapping for X, and
2. when the LDP connection between Ru and Rd was opened, Ru
indicated that it could support the POP label, and
3. Rd is an LSP Egress (not proxy egress) for X.
This causes the penultimate LSR on a LSP to pop the label stack. This
is quite appropriate; if the LSP Egress is an MPLS Egress for X, then
if the penultimate LSR does not pop the label stack, the LSP Egress
will need to look up the label, pop the label stack, and then look up
the next label (or look up the L3 address, if no more labels are
present). By having the penultimate LSR pop the label stack, the LSP
Egress is saved the work of having to look up two labels in order to
make its forwarding decision.
Rosen, Viswanathan & Callon [Page 39]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
However, if the penultimate LSR is an ATM switch, it may not have the
capability to pop the label stack. Hence a POP label mapping may be
distributed only to LSRs which can support that function.
If the penultimate LSR in an LSP for address prefix X is an LSP Proxy
Egress, it acts just as if the LSP Egress had distributed the POP
label for X.
3.1.6. Option: Egress-Targeted Label Assignment
There are situations in which an LSP Ingress, Ri, knows that packets
of several different Streams must all follow the same LSP,
terminating at, say, LSP Egress Re. In this case, proper routing can
be achieved by using a single label can be used for all such Streams;
it is not necessary to have a distinct label for each Stream. If
(and only if) the following conditions hold:
1. the address of LSR Re is itself in the routing table as a "host
route", and
2. there is some way for Ri to determine that Re is the LSP egress
for all packets in a particular set of Streams
Then Ri may bind a single label to all FECS in the set. This is
known as "Egress-Targeted Label Assignment."
How can LSR Ri determine that an LSR Re is the LSP Egress for all
packets in a particular Stream? There are a couple of possible ways:
- If the network is running a link state routing algorithm, and all
nodes in the area support MPLS, then the routing algorithm
provides Ri with enough information to determine the routers
through which packets in that Stream must leave the routing
domain or area.
- It is possible to use LDP to pass information about which address
prefixes are "attached" to which egress LSRs. This method has
the advantage of not depending on the presence of link state
routing.
If egress-targeted label assignment is used, the number of labels
that need to be supported throughout the network may be greatly
reduced. This may be significant if one is using legacy switching
hardware to do MPLS, and the switching hardware can support only a
limited number of labels.
One possible approach would be to configure the network to use
Rosen, Viswanathan & Callon [Page 40]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
egress-targeted label assignment by default, but to configure
particular LSRs to NOT use egress-targeted label assignment for one
or more of the address prefixes for which it is an LSP egress. We
impose the following rule:
- If a particular LSR is NOT an LSP Egress for some set of address
prefixes, then it should assign labels to the address prefixes in
the same way as is done by its LSP next hop for those address
prefixes. That is, suppose Rd is Ru's LSP next hop for address
prefixes X1 and X2. If Rd assigns the same label to X1 and X2,
Ru should as well. If Rd assigns different labels to X1 and X2,
then Ru should as well.
For example, suppose one wants to make egress-targeted label
assignment the default, but to assign distinct labels to those
address prefixes for which there are multiple possible LSP egresses
(i.e., for those address prefixes which are multi-homed.) One can
configure all LSRs to use egress-targeted label assignment, and then
configure a handful of LSRs to assign distinct labels to those
address prefixes which are multi-homed. For a particular multi-homed
address prefix X, one would only need to configure this in LSRs which
are either LSP Egresses or LSP Proxy Egresses for X.
It is important to note that if Ru and Rd are adjacent LSRs in an LSP
for X1 and X2, forwarding will still be done correctly if Ru assigns
distinct labels to X1 and X2 while Rd assigns just one label to the
both of them. This just means that R1 will map different incoming
labels to the same outgoing label, an ordinary occurrence.
Similarly, if Rd assigns distinct labels to X1 and X2, but Ru assigns
to them both the label corresponding to the address of their LSP
Egress or Proxy Egress, forwarding will still be done correctly. Ru
will just map the incoming label to the label which Rd has assigned
to the address of that LSP Egress.
3.2. MPLS and Explicitly Routed LSPs
There are a number of reasons why it may be desirable to use explicit
routing instead of hop by hop routing. For example, this allows
routes to be based on administrative policies, and allows the routes
that LSPs take to be carefully designed to allow traffic engineering
(i.e., to allow intentional management of the loading of the
bandwidth through the nodes and links in the network).
Rosen, Viswanathan & Callon [Page 41]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
3.2.1. Explicitly Routed LSP Tunnels: Traffic Engineering
In some situations, the network administrators may desire to forward
certain classes of traffic along certain pre-specified paths, where
these paths differ from the Hop-by-hop path that the traffic would
ordinarily follow. This is known as Traffic Engineering.
MPLS allows this to be easily done by means of Explicitly Routed LSP
Tunnels. All that is needed is:
1. A means of selecting the packets that are to be sent into the
Explicitly Routed LSP Tunnel;
2. A means of setting up the Explicitly Routed LSP Tunnel;
3. A means of ensuring that packets sent into the Tunnel will not
loop from the receive endpoint back to the transmit endpoint.
If the transmit endpoint of the tunnel wishes to put a labeled packet
into the tunnel, it must first replace the label value at the top of
the stack with a label value that was distributed to it by the
tunnel's receive endpoint. Then it must push on the label which
corresponds to the tunnel itself, as distributed to it by the next
hop along the tunnel. To allow this, the tunnel endpoints should be
explicit LDP peers. The label mappings they need to exchange are of
no interest to the LSRs along the tunnel.
3.3. Label Stacks and Implicit Peering
Suppose a particular LSR Re is an LSP proxy egress for 10 address
prefixes, and it reaches each address prefix through a distinct
interface.
One could assign a single label to all 10 address prefixes. Then Re
is an LSP egress for all 10 address prefixes. This ensures that
packets for all 10 address prefixes get delivered to Re. However, Re
would then have to look up the network layer address of each such
packet in order to choose the proper interface to send the packet on.
Alternatively, one could assign a distinct label to each interface.
Then Re is an LSP proxy egress for the 10 address prefixes. This
eliminates the need for Re to look up the network layer addresses in
order to forward the packets. However, it can result in the use of a
large number of labels.
An alternative would be to bind all 10 address prefixes to the same
level 1 label (which is also bound to the address of the LSR itself),
Rosen, Viswanathan & Callon [Page 42]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
and then to bind each address prefix to a distinct level 2 label. The
level 2 label would be treated as an attribute of the level 1 label
mapping, which we call the "Stack Attribute". We impose the
following rules:
- When LSR Ru initially labels an untagged packet, if the longest
match for the packet's destination address is X, and R's LSP next
hop for X is Rd, and Rd has distributed to R1 a mapping of label
L1 X, along with a stack attribute of L2, then
1. Ru must push L2 and then L1 onto the packet's label stack,
and then forward the packet to Rd;
2. When Ru distributes label mappings for X to its LDP peers,
it must include L2 as the stack attribute.
3. Whenever the stack attribute changes (possibly as a result
of a change in Ru's LSP next hop for X), Ru must distribute
the new stack attribute.
Note that although the label value bound to X may be different at
each hop along the LSP, the stack attribute value is passed
unchanged, and is set by the LSP proxy egress.
Thus the LSP proxy egress for X becomes an "implicit peer" with each
other LSR in the routing area or domain. In this case, explicit
peering would be too unwieldy, because the number of peers would
become too large.
3.4. MPLS and Multi-Path Routing
If an LSR supports multiple routes for a particular Stream, then it
may assign multiple labels to the Stream, one for each route. Thus
the reception of a second label mapping from a particular neighbor
for a particular address prefix should be taken as meaning that
either label can be used to represent that address prefix.
If multiple label mappings for a particular address prefix are
specified, they may have distinct attributes.
Rosen, Viswanathan & Callon [Page 43]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
3.5. LSPs may be Multipoint-to-Point Entities
Consider the case of packets P1 and P2, each of which has a
destination address whose longest match, throughout a particular
routing domain, is address prefix X. Suppose that the Hop-by-hop
path for P1 is , and the Hop-by-hop path for P2 is . Let's suppose that R3 binds label L3 to X, and distributes
this mapping to R2. R2 binds label L2 to X, and distributes this
mapping to both R1 and R4. When R2 receives packet P1, its incoming
label will be L2. R2 will overwrite L2 with L3, and send P1 to R3.
When R2 receives packet P2, its incoming label will also be L2. R2
again overwrites L2 with L3, and send P2 on to R3.
Note then that when P1 and P2 are traveling from R2 to R3, they carry
the same label, and as far as MPLS is concerned, they cannot be
distinguished. Thus instead of talking about two distinct LSPs, and , we might talk of a single "Multipoint-to-
Point LSP", which we might denote as <{R1, R4}, R2, R3>.
This creates a difficulty when we attempt to use conventional ATM
switches as LSRs. Since conventional ATM switches do not support
multipoint-to-point connections, there must be procedures to ensure
that each LSP is realized as a point-to-point VC. However, if ATM
switches which do support multipoint-to-point VCs are in use, then
the LSPs can be most efficiently realized as multipoint-to-point VCs.
Alternatively, if the SVP Multipoint Encoding (section 2.21) can be
used, the LSPs can be realized as multipoint-to-point SVPs.
3.6. LSP Tunneling between BGP Border Routers
Consider the case of an Autonomous System, A, which carries transit
traffic between other Autonomous Systems. Autonomous System A will
have a number of BGP Border Routers, and a mesh of BGP connections
among them, over which BGP routes are distributed. In many such
cases, it is desirable to avoid distributing the BGP routes to
routers which are not BGP Border Routers. If this can be avoided,
the "route distribution load" on those routers is significantly
reduced. However, there must be some means of ensuring that the
transit traffic will be delivered from Border Router to Border Router
by the interior routers.
This can easily be done by means of LSP Tunnels. Suppose that BGP
routes are distributed only to BGP Border Routers, and not to the
interior routers that lie along the Hop-by-hop path from Border
Router to Border Router. LSP Tunnels can then be used as follows:
Rosen, Viswanathan & Callon [Page 44]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
1. Each BGP Border Router distributes, to every other BGP Border
Router in the same Autonomous System, a label for each address
prefix that it distributes to that router via BGP.
2. The IGP for the Autonomous System maintains a host route for
each BGP Border Router. Each interior router distributes its
labels for these host routes to each of its IGP neighbors.
3. Suppose that:
a) BGP Border Router B1 receives an unlabeled packet P,
b) address prefix X in B1's routing table is the longest
match for the destination address of P,
c) the route to X is a BGP route,
d) the BGP Next Hop for X is B2,
e) B2 has bound label L1 to X, and has distributed this
mapping to B1,
f) the IGP next hop for the address of B2 is I1,
g) the address of B2 is in B1's and I1's IGP routing tables
as a host route, and
h) I1 has bound label L2 to the address of B2, and
distributed this mapping to B1.
Then before sending packet P to I1, B1 must create a label
stack for P, then push on label L1, and then push on label L2.
4. Suppose that BGP Border Router B1 receives a labeled Packet P,
where the label on the top of the label stack corresponds to an
address prefix, X, to which the route is a BGP route, and that
conditions 3b, 3c, 3d, and 3e all hold. Then before sending
packet P to I1, B1 must replace the label at the top of the
label stack with L1, and then push on label L2.
With these procedures, a given packet P follows a level 1 LSP all of
whose members are BGP Border Routers, and between each pair of BGP
Border Routers in the level 1 LSP, it follows a level 2 LSP.
These procedures effectively create a Hop-by-Hop Routed LSP Tunnel
between the BGP Border Routers.
Since the BGP border routers are exchanging label mappings for
Rosen, Viswanathan & Callon [Page 45]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
address prefixes that are not even known to the IGP routing, the BGP
routers should become explicit LDP peers with each other.
3.7. Other Uses of Hop-by-Hop Routed LSP Tunnels
The use of Hop-by-Hop Routed LSP Tunnels is not restricted to tunnels
between BGP Next Hops. Any situation in which one might otherwise
have used an encapsulation tunnel is one in which it is appropriate
to use a Hop-by-Hop Routed LSP Tunnel. Instead of encapsulating the
packet with a new header whose destination address is the address of
the tunnel's receive endpoint, the label corresponding to the address
prefix which is the longest match for the address of the tunnel's
receive endpoint is pushed on the packet's label stack. The packet
which is sent into the tunnel may or may not already be labeled.
If the transmit endpoint of the tunnel wishes to put a labeled packet
into the tunnel, it must first replace the label value at the top of
the stack with a label value that was distributed to it by the
tunnel's receive endpoint. Then it must push on the label which
corresponds to the tunnel itself, as distributed to it by the next
hop along the tunnel. To allow this, the tunnel endpoints should be
explicit LDP peers. The label mappings they need to exchange are of
no interest to the LSRs along the tunnel.
3.8. MPLS and Multicast
Multicast routing proceeds by constructing multicast trees. The tree
along which a particular multicast packet must get forwarded depends
in general on the packet's source address and its destination
address. Whenever a particular LSR is a node in a particular
multicast tree, it binds a label to that tree. It then distributes
that mapping to its parent on the multicast tree. (If the node in
question is on a LAN, and has siblings on that LAN, it must also
distribute the mapping to its siblings. This allows the parent to
use a single label value when multicasting to all children on the
LAN.)
When a multicast labeled packet arrives, the NHLFE corresponding to
the label indicates the set of output interfaces for that packet, as
well as the outgoing label. If the same label encoding technique is
used on all the outgoing interfaces, the very same packet can be sent
to all the children.
Rosen, Viswanathan & Callon [Page 46]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
4. LDP Procedures
This section is FFS.
5. Security Considerations
Security considerations are not discussed in this version of this
draft.
6. Authors' Addresses
Eric C. Rosen
Cisco Systems, Inc.
250 Apollo Drive
Chelmsford, MA, 01824
E-mail: erosen@cisco.com
Arun Viswanathan
IBM Corp.
17 Skyline Drive
Hawthorne NY 10532
914-784-3273
E-mail: arunv@vnet.ibm.com
Ross Callon
Ascend Communications, Inc.
1 Robbins Road
Westford, MA 01886
508-952-7412
E-mail: rcallon@casc.com
7. References
[1] "A Framework for Multiprotocol Label Switching", R.Callon,
P.Doolan, N.Feldman, A.Fredette, G.Swallow, and A.Viswanathan, work
in progress, Internet Draft <draft-ietf-mpls-framework-01.txt>, July
1997.
[2] "ARIS: Aggregate Route-Based IP Switching", A. Viswanathan, N.
Feldman, R. Boivie, R. Woundy, work in progress, Internet Draft
<draft-viswanathan-aris-overview-00.txt>, March 1997.
[3] "ARIS Specification", N. Feldman, A. Viswanathan, work in
progress, Internet Draft <draft-feldman-aris-spec-00.txt>, March
1997.
Rosen, Viswanathan & Callon [Page 47]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
[4] "ARIS Support for LAN Media Switching", S. Blake, A. Ghanwani, W.
Pace, V. Srinivasan, work in progress, Internet Draft , March 1997.
[5] "Tag Switching Architecture - Overview", Rekhter, Davie, Katz,
Rosen, Swallow, Farinacci, work in progress, Internet Draft , January, 1997.
[6] "Tag distribution Protocol", Doolan, Davie, Katz, Rekhter, Rosen,
work in progress, Internet Draft <draft-doolan-tdp-spec-01.txt>, May,
1997.
[7] "Use of Tag Switching with ATM", Davie, Doolan, Lawrence,
McGloghrie, Rekhter, Rosen, Swallow, work in progress, Internet Draft
<draft-davie-tag-switching-atm-01.txt>, January, 1997.
[8] "Label Switching: Label Stack Encodings", Rosen, Rekhter, Tappan,
Farinacci, Fedorkow, Li, work in progress, Internet Draft , June, 1997.
[9] "Partitioning Tag Space among Multicast Routers on a Common
Subnet", Farinacci, work in progress, internet draft , December, 1996.
[10] "Multicast Tag Binding and Distribution using PIM", Farinacci,
Rekhter, work in progress, internet draft , December, 1996.
[11] "Toshiba's Router Architecture Extensions for ATM: Overview",
Katsube, Nagami, Esaki, RFC 2098, February, 1997.
[12] "Loop-Free Routing Using Diffusing Computations", J.J. Garcia-
Luna-Aceves, IEEE/ACM Transactions on Networking, Vol. 1, No. 1,
February 1993.
Appendix A Why Egress Control is Better
This section is written by Arun Viswanathan.
It is demonstrated here why egress control is a necessary and
sufficient mechanism for the LDP, and therefore is the optimal method
for setting up LSPs.
The necessary condition is established by citing counter examples
that can be achieved *only* by egress control. It's also established
why these typical scenarios are vital requirements for a
multiprotocol LDP. The sufficiency part is established by proving
Rosen, Viswanathan & Callon [Page 48]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
that egress control subsumes the local control.
Then finally, some discussions are made to mitigate concerns
expressed against not having local control. It is shown that local
control has clearly undesirable properties which may lead to severe
scalability and robustness problems. It is also shown that in having
both egress control and local control simultaneously in a network
leads to interoperability problems and how local control abrogates
the essential benefits of egress control.
A complete and self-contained case is presented here that clearly
establishes that egress control is the preponderant mechanism for
LDP, and it suffices to support egress control alone as the
distribution paradigm.
A.1 Definition of an Egress
A node is identified as an "egress" for a Stream, if:
1) it's at a routing boundary for that Stream,
2) the next hop for that Stream is non-MPLS,
3) the Stream is directly attached or the node itself.
Nodes that satisfy conditions 1 or 2 for Streams, will by default
start behaving as egress for those streams. Note that conditions 1
and 2 can be learned dynamically. For condition 3, nodes will not by
default act as an egress for themselves or directly attached
networks. If this condition is made the default, the LSPs setup by
egress control will create LSPs that are identical to the LSPs
created by local control.
A.2 Overview of Egress Control
When a node is an egress for a Stream, it originates a LSP setup
message for that particular Stream. The setup message is sent to all
MPLS neighbors, except the next hop neighbor. Each of these messages
to the neighbors carry an appropriate label for that Stream. When a
node in a MPLS domain receives a setup message from a neighbor for a
particular Stream, it checks if that neighbor is the next hop for the
given Stream. If so, it propagates the message to all its MPLS
neighbors, except the next hop from which the message arrived. If
not, the node may keep the label provided in the setup message for
future use or negatively acknowledge the node that sent the message
to release the label assignment. But it must not forward the setup
message from the incorrect next hop to any of its neighbors. This
flooding scheme is similar in mechanism to Reverse Path Multicast.
When a next hop for a Stream changes due to change in network
Rosen, Viswanathan & Callon [Page 49]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
topology, or a new node joins the topology, the node is locally
appended to the existing LSP, without requiring egress intervention.
The node may either request the label mapping from the new next hop,
or use the previously stored (but unused) label from that next hop.
In the former case, the new next hop immediately responds with a
label mapping for that Stream if it has its own downstream mapping
for that Stream.
A.3 Why Egress Control is Necessary
There are some important situations in which egress control is
necessary:
- Shutting off an LSP
If for some reason a network administrator requires to "shut off"
a LSP setup for a particular Stream, s/he can configure the
egress node for that Stream for the desired result. Note that
the requirement to shut off an LSP is a very fundamental one. If
a destination has network layer reachability but no MPLS layer
reachability (because of a problem in MPLS layer), shutting off
an LSP provides the only means to reach that destination. This
mode of operation can be used by LSRs in a network that aren't a
sink for large amounts of data. These LSRs usually require an
occasional telnet or network management traffic. It's important
to provide the capability that such nodes in a network can be
accessed through hop-by-hop connectivity avoiding the MPLS layer
optimization. The reachability is more important than
optimization in instances like this. The MPLS architecture MUST
provide this capability.
Note that this is only possible in local control when each node
in an entire network is configured to shut off a LSP setup for a
particular Stream. Such is neither desirable nor scalable.
- Egress Aggregation
In some networks, due to the absence of routing summarization,
aggregation may not be possible through routing information.
However, with Egress control, it is possible to aggregate *all*
Streams that exit the network through a common egress node with a
single LSP. This is achieved easily because the egress simply
can use the same label for all Streams.
Such is simply not possible with the Local control; with local
knowledge LSRs cannot map several Streams to a single label
because it is unknown if Streams will diverge at some subsequent
downstream node.
Rosen, Viswanathan & Callon [Page 50]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
The egress aggregation works for both distance vector protocols
and link state protocols; it is protocol independent. Note that
when using VP switching in conjunction with some distance vector
protocols it becomes very essential that such aggregation be
possible, as there are many vendor switches that don't have VC
merging capability, and have limited VP switching capability.
The egress control provides such vendors with a level-playing
field to compete with MPLS products. Moreover, this capability
can be very useful in enterprise networks; where several legacy
LANs at a site can be aggregated to the egress LSR at that site.
Furthermore, this approach can drastically reduce signalling and
LSP state maintenance overheads in the entire network.
- Loop Prevention
The loop-prevention mechanism only works from the egress node for
multipoint-to-point LSPs, since the loop prevention mechanism
requires the list of LSR nodes through which the setup message
has already traversed in order to identify and prevent LSP loops.
A loop prevention scheme is not possible through local control.
- De-aggregation
Egress control provides the capability to de-aggregate one or
more Streams from an aggregated Stream. For example, if a
network is aggregating all CIDRs of an EBGP node into a single
LSP, with egress control, a specific CIDR from this bundle can be
given its own dedicated LSP. This enables one to apply special
policies to specific CIDRs when required.
In the local control this can be achieved only by configuring
every node in the network with specific de-aggregation
information and the associated policy. This approach can lead
severe scalability problems.
- Unique Labels
As is known, when using VP merging, all ingresses must have
unique VCI values to prevent cell interleaving. With egress
control, it is possible to distribute unique VCI values to the
ingress nodes, avoiding the need to configure each ingress node.
The egress node can pick a unique VCI for each ingress node.
Another benefit of egress control is that each egress can be
configured with a unique label value in the case of egress
aggregation (as described above). Since the label value is
unique, the same label value can be used on all the segments of a
LSP. This enables one to identify anywhere in a network each LSP
Rosen, Viswanathan & Callon [Page 51]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
that is associated with a certain egress node, thus easing
network debugging.
This again, is not possible in the local control because of the
lack of a single coordinating node.
A.4 Examples that work better through egress control
Local control needs to propagate attributes that come from the
downstream node to all upstream nodes. This behavior itself can be
LIKENED to the egress control. Nevertheless, the local control can
achieve these only in a severely inefficient manner. Since each node
only knows of local information, it creates and distributes an LSP
with incorrect attributes. As each node learns of new downstream
attributes, a correction is made as the attributes are propagated
upstream again. This can lead to a worst case of O(n-squared) setup
messages to create a single LSP, where n is the number of nodes in a
LSP.
In the egress control, the attribute distribution is achieved during
initial LSP setup, with a single message from the egress to
ingresses.
- TTL/Traceroute
The ingress requires a proper LSP hop-count value to decrement
TTL in packets that use a particular LSP, in environments such as
ATM which do not have a TTL equivalent. This simulates the TTL
decrement which exists in an IP network, and also enables scoping
utilities, such as traceroute, to work as they do today in IP
networks. In egress control, the LSP hop-count is known at the
ingress as a by-product of the LSP setup message, since an LSP
setup message traverses from egress to ingress, and increments
the hop-count at each node along the path.
- MTU
When the MTU at the egress node is smaller than the MTU at some
of the ingress nodes, packets originated at those ingress nodes
will be dropped when they reach the egress node. Hosts not using
MTU discovery have no means to recover from this. However,
similar to the hop-count, the minimum LSP MTU can be propagated
to the ingresses via egress control LSP setup messages, enabling
the ingress to do fragmentation when required.
Rosen, Viswanathan & Callon [Page 52]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
- Implicit Peering
Implicit peering is the mechanism through which higher level
stack labels are communicated to the ingress nodes. These label
values are piggybacked in the LSP setup messages. This works
best with egress control; when the egress creates the setup
message, it can piggyback the stack labels at the same time.
- ToS/COS Based LSPs
When certain LSPs require higher or lower precedence or priority
through a network, the single egress node for that LSP can be
configured with the required priority and this can be
communicated in the egress control LSP setup message. In the
local control, each and every node in the network must be
configured per LSP to achieve the same result.
The local control initially distributes labels to its neighbors
willy-nilly, and then waits for attributes to come through egress
control. Thus, local control is completely dependent on egress
control to provide complete functional operation to LSPs. Otherwise,
local control requires that attributes be configured through the
entire network for each Stream. This is the most compelling argument
that local control is *not sufficient*; or conversely, egress control
is necessary. This demonstrates egress control subsumes the local
control. Moreover, distribution of labels without associated
attributes may not be appropriate and may lead to undesired results.
A.5 Egress Control is Sufficient
The argument for sufficiency is proved by demonstrating that required
LSPs can be created with egress control, and this is not the case
with local control.
The egress control can create an LSP for every route entry made by
the routing protocols:
1. A route can be learned from another routing domain, in which
case the LSR at the routing domain will act as an egress for
the route and originate an LSP setup for that route.
2. A route can be a locally attached network or the LSR itself may
be a host route. In this case, the LSR to which such a route
is attached originates an LSP setup message.
Rosen, Viswanathan & Callon [Page 53]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
3. An LSR with a non-MPLS next-hop behaves as an egress for all
those route whose next-hop is the non-MPLS neighbor.
These three above methods can create an LSP for each route entry in a
network. Moreover, policy specific LSPs, as described previously,
can *only* be achieved with egress control. Thus, egress control is
necessary and sufficient for creating LSPs. QED.
A.6 Discussions
A.6.1 Is Local control faster than Egress control?
During topology changes, such as links going down, coming up, change
in link cost, etc, there is no difference in setup latency between
Egress Control and Local control. This is due to the fact that the
node (Ru) which undergoes a change in next-hop for a Stream
immediately requests a label assignment from the new next hop node
(Rd). The new next hop node then immediately supplies the label
mapping for the requested Stream. As explained in the Egress Control
Method section, the node Ru may already have stored label assignments
from the node Rd, in which case node Ru can immediately splice itself
to the multipoint-to-point tree. Hence, new nodes are spliced into
existing LSPs locally. In the scenario where a network initially
learns of a new route, although the Local control may setup LSPs
faster than the Egress control, this difference in latency has no
perceived advantage. Since routing itself may take several seconds
to propagate and converge on the new route information, the potential
latency of egress control is small as compared to the routing
protocol propagation time, and the initial setup time at route
propagation time is unimportant since these are long lived LSPs.
Moreover, the hurried distribution of labels in local control may not
carry much meaning because:
4. The associated attributes are not applied or propagated to the
ingress.
5. While the ingress may believe it has an LSP, in reality the
packets may be blackholed in the middle of the network if the
full LSP is not established.
6. Policy based LSPs, which can only be achieved via egress
control as described above, may undo an un-used label
assignment established by local control.
A.6.2 Scalability and Robustness
It has been alleged that the egress control does not have the
Rosen, Viswanathan & Callon [Page 54]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
scalability and robustness properties required by distributed
processing. However, the egress uses a root distribution paradigm
commonly used by many other standard routing protocols. For example,
in the case of OSPF, LSAs are flooded through a domain originating at
the "egress", where the difference being that the flooding in the
case of OSPF is contained through a sequence number and in the Egress
control it is contained by the next hop validation. In the case of
PIM (and some other multicast protocols), the distribution mechanism
is in fact exactly similar. Even in BGP with route reflection,
updates originate at the root and traverse a tree structure to reach
the peers, as opposed to a n-square mesh. The commonality is the
distribution paradigm, in which the distribution originates at the
root of a tree and traverses the branches till it reaches all the
leaves. None of the above mentioned protocols have scalability or
robustness problems because of the distribution paradigm.
The ONLY concern expressed against to counter Egress control is that
if the setup message does not propagate upstream from a certain node,
then the sub-tree upstream of that node will not be added into the
LSP. It's a reasonable concern, but further analysis shows that it's
not a realistic problem. The impact of this problem compared to the
impact of a similar problem in local control are exactly the same
when LSRs employed in a MPLS domain have little or no forwarding
capabilities (for example, ATM LSRs), since in both cases, packets
are blackholed. In fact, in the egress control the packets for
afflicted LSPs will be dropped right at the ingress, while with local
control the packets will be dropped at the point of breakage, causing
packets to unnecessarily traverse part way through the network. When
reasonable forwarding capability exists in the MPLS domain, with the
egress control the packets may be forwarded hop-by-hop till the point
where the LSP setup ended. Whereas in case of local control, the
packets will label switched till the point of breakage and hop-by-hop
forwarded till the LSP segment resumes. Since egress control has
advantages when there is no forwarding capability, and local control
is has advantages when there is forwarding capability, there is an
equal tradeoff between them, and thus, neither is superior or
inferior in this regard. This latter case is simply a loss in
optimization, since the network has reasonable forwarding
capabilities. Hence the robustness issue is not a problem in either
types of networks. As mentioned before, the local control is
dependent on egress control for distributing attributes. The
attribute distribution could then also face the same problem of
stalled propagation, which would lead to erroneous LSP setup. So,
the local control can also be seen as afflicted with this problem, if
it exists.
Moreover, if stalled propagation were truly a problem, there are
other schemes in MPLS that would face the same issue. For example,
Rosen, Viswanathan & Callon [Page 55]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
the label distribution through PIM, Explicit Route setup, and RSVP
would also not work, and therefore should be withdrawn :-).
Note that exhaustion of label space cannot stall the propagation of
messages to the upstream nodes. Appropriate indications can be given
to the upstream nodes in the setup message that no label allocation
was made because of exhaustion of label space, so that correct action
can be taken at the upstream nodes, and yet the LSP setup would
continue.
A.6.3 Conclusion
The attempt here is not to deride the local control, but since one
method subsumes the features and properties of the other, then why
support both and complicate implementation, interoperability and
maintenance? In fact RFC1925 says, "In protocol design, perfection
has been reached not when there is nothing left to add, but when
there is nothing left to take away". A usual diplomatic resolution
for such controversy is to make accommodations for both. We feel
that it's a poor choice of architecture to support both. That is why
we feel strongly that this must be evaluated by the MPLS WG.
In a way, controlling the network behavior as to which LSP are
formed, which Streams map to which LSPs, and the associated
attributes, can be compared to applying policies at the edges of an
AS. This is precisely what the egress control provides, a rich and
varied policy control at the egress node of LSPs.
Appendix B Why Local Control is Better
This section is written by Eric Rosen.
The remaining area of dispute between advocates of "local control"
and advocates of "egress control" is relatively small. In
particular, there is agreement on the following points:
1. If LSR R1's next hop for address prefix X is LSR R2, and R2 is
in a different area or in a different routing domain than R1,
then R1 may assign and distribute a label for X, even if R2 has
not done so.
This means that even under egress control, the border routers
in one autonomous system do not have to wait, before
distributing labels, for any downstream routers which are in
other autonomous systems.
Rosen, Viswanathan & Callon [Page 56]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
2. If LSR R1's next hop for address prefix X is LSR R2, but R1
receives a label mapping for X from LSR R3, then R1 may
remember R3's mapping. If, at some later time, R3 becomes R1's
next hop for S, then (if R1 is not using loop prevention) R1
may immediately begin using R3 as the LSP next hop for S, using
the remembered mapping from R3.
3. Attributes which are passed upstream from the egress may change
over time, as a result of reconfiguration of the egress, or of
other events. This means that even if egress control is used,
LSRs must be able to accept attribute changes on existing LSPs;
attributes are not fixed when the LSP is first constructed, nor
does a change in attributes require a new LSP to be
constructed.
The dispute is centered on the situation in which the following
conditions hold:
- LSR R1's next hop for address prefix X is within the same
administrative domain as R1, and
- R1's next hop for X has not distributed to R1 a label for X, and
- R1 has not yet distributed to its neighbors any labels for X.
With local control, R1 is permitted to distribute a label for X to
its neighbors; with egress control it is not.
From an implementation perspective, the difference then between
egress control and local control is relatively small. Egress control
simply creates an additional state in the label distribution process,
and prohibits label distribution in that state.
From the perspective of network behavior, however, this difference is
a bit more significant:
- Egress control adds latency to the initial construction of an
LSP, because the path must be set up serially, node by node from
the egress. With local control, all LSRs along the path may
perform their setup activities in parallel.
- Egress control adds additional interdependencies among nodes, as
there is something that one node cannot do until some other node
does something else first, which it cannot do until some other
node does something first, etc. This is problematical for a
number of reasons.
Rosen, Viswanathan & Callon [Page 57]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
* In robust system design, one tries to avoid such
interdependencies, since they always bring along robustness
and scalability problems.
* In some situations, it is advantageous for a node to use
MPLS, even if some node downstream is not functioning
properly and hence not assigning labels as it should.
These disadvantages might be tolerable if there is some significant
problem which can be solved by egress control, but not by local
control. So it is worth looking to see if there is such a problem.
There are a number of situations in which it may be desirable for an
LSP Ingress node to know certain attributes of the LSP, e.g., the
number of hops in the LSP. It is sometimes claimed that obtaining
such information requires the use of egress control. However, this
is not true. Any attribute of an LSP is liable to change after the
LSP exists. Procedures to detect and communicate the change must
exist. These procedures CANNOT be tied to the initial construction
of the LSP, since they must execute after the LSP has already been
constructed. The ability to pass control information upstream along
a path towards an ingress node does not presuppose anything about the
procedures used to construct the path.
The fundamental issue separating the advocates of egress control from
the advocates of local control is really a network management issue.
To advocates of egress control, setting up an LSP for a particular
address prefix is analogous to setting up a PVC in an ATM network.
When setting up a PVC, one goes to one of the PVC endpoints and
enters certain configuration information. Similarly, one might think
that to set up an LSP for a particular address prefix, one goes to
the LSR which is the egress for that address prefix, and enters
configuration information. This allows the network administrator
complete control of which address prefixes are assigned LSPs and
which are not. And if this is one's management model, egress control
does simplify the configuration issues.
On the other hand, if one's model is that the LSPs get set up
automatically by the network, as a result of the operation of the
routing algorithm, then egress control is of no utility at all. When
one hears the claim that "egress control allow you to control your
network from a few nodes", what is really being claimed is "egress
control simplifies the job of manually configuring all the LSPs in
your network". Of course, if you don't intend to manually configure
all the LSPs in your network, this is irrelevant.
So before an egress control scheme is adopted, one should ask whether
complete manual configuration of the set of address prefixes which
Rosen, Viswanathan & Callon [Page 58]
Internet Draft draft-ietf-mpls-arch-00.txt August 1997
get assigned LSPs is necessary. That is, is this capability needed
to solve a real problem?
It is sometimes claimed that egress control is needed if one wants to
conserve labels by assigning a single label to all address prefixes
which have the same egress. This is not true. If the network is
running a link state routing algorithm, each LSR already knows which
address prefixes have a common egress, and hence can assign a common
label. If the network is running a distance vector routing protocol,
information about which address prefixes have a common egress can be
made to "bubble up" from the egress, using LDP, even if local control
is used.
It is only in the case where the number of available labels is so
small that their use must be manually administered that egress
control has an advantage. It may be arguable that egress control
should be an option that can be used for the special cases in which
it provides value. In most cases, there is no reason to have it at
all.
Rosen, Viswanathan & Callon [Page 59]