Request for Comments: 3618 D. Meyer, Ed.
Category: Experimental October 2003
Multicast Source Discovery Protocol (MSDP)
Status of this Memo
This memo defines an Experimental Protocol for the Internet
community. It does not specify an Internet standard of any kind.
Discussion and suggestions for improvement are requested.
Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2003). All Rights Reserved.
Abstract
The Multicast Source Discovery Protocol (MSDP) describes a mechanism
to connect multiple IP Version 4 Protocol Independent Multicast
Sparse-Mode (PIM-SM) domains together. Each PIM-SM domain uses its
own independent Rendezvous Point (RP) and does not have to depend on
RPs in other domains. This document reflects existing MSDP
implementations.
Table of Contents
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
5. Timers. . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
5.1. SA-Advertisement-Timer . . . . . . . . . . . . . . . . . 5
5.2. SA-Advertisement-Timer Processing. . . . . . . . . . . . 5
5.3. SA Cache Timeout (SA-State Timer). . . . . . . . . . . . 5
5.4. Peer Hold Timer. . . . . . . . . . . . . . . . . . . . . 5
5.5. KeepAlive Timer. . . . . . . . . . . . . . . . . . . . . 6
5.6. ConnectRetry Timer . . . . . . . . . . . . . . . . . . . 6
6. Intermediate MSDP Peers . . . . . . . . . . . . . . . . . . . 6
7. SA Filtering and Policy . . . . . . . . . . . . . . . . . . . 6
8. Encapsulated Data Packets . . . . . . . . . . . . . . . . . . 7
9. Other Scenarios . . . . . . . . . . . . . . . . . . . . . . . 7
10. MSDP Peer-RPF Forwarding. . . . . . . . . . . . . . . . . . . 7
10.1. Definitions . . . . . . . . . . . . . . . . . . . . . . 7
10.1.1. Multicast RPF Routing Information Base. . . . . 8
10.1.2. Peer-RPF Route. . . . . . . . . . . . . . . . . 8
10.1.3. Peer-RPF Forwarding Rules . . . . . . . . . . . 8
10.2. MSDP mesh-group semantics . . . . . . . . . . . . . . . 9
11. MSDP Connection State Machine . . . . . . . . . . . . . . . . 9
11.1. Events. . . . . . . . . . . . . . . . . . . . . . . . . 10
11.2. Actions . . . . . . . . . . . . . . . . . . . . . . . . 10
11.3. Peer-specific Events. . . . . . . . . . . . . . . . . . 11
11.4. Peer-independent Events . . . . . . . . . . . . . . . . 11
12. Packet Formats. . . . . . . . . . . . . . . . . . . . . . . . 12
12.1. MSDP TLV format . . . . . . . . . . . . . . . . . . . . 12
12.2. Defined TLVs. . . . . . . . . . . . . . . . . . . . . . 12
12.2.1. IPv4 Source-Active TLV. . . . . . . . . . . . . 13
12.2.2. KeepAlive TLV . . . . . . . . . . . . . . . . . 14
13. MSDP Error Handling . . . . . . . . . . . . . . . . . . . . . 15
14. SA Data Encapsulation . . . . . . . . . . . . . . . . . . . . 15
15. Applicability Statement . . . . . . . . . . . . . . . . . . . 15
15.1. Between PIM Domains . . . . . . . . . . . . . . . . . . 15
15.2. Between Anycast-RPs . . . . . . . . . . . . . . . . . . 15
16. Intellectual Property . . . . . . . . . . . . . . . . . . . . 15
17. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 16
18. Security Considerations . . . . . . . . . . . . . . . . . . . 16
19. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 17
19.1. Allocated TLV Range . . . . . . . . . . . . . . . . . . 17
19.2. Experimental TLV Range. . . . . . . . . . . . . . . . . 17
20. References. . . . . . . . . . . . . . . . . . . . . . . . . . 17
20.1. Normative References. . . . . . . . . . . . . . . . . . 17
20.2. Informative References. . . . . . . . . . . . . . . . . 18
21. Editors’ Addresses. . . . . . . . . . . . . . . . . . . . . . 18
22. Full Copyright Statement. . . . . . . . . . . . . . . . . . . 19
1. Introduction
The Multicast Source Discovery Protocol (MSDP) describes a mechanism
to connect multiple PIM Sparse-Mode (PIM-SM) [RFC2362] domains
together. Each PIM-SM domain uses its own independent RP(s) and does
not have to depend on RPs in other domains. Advantages of this
approach include:
o No Third-party resource dependencies on a domain’s RP
PIM-SM domains can rely on their own RPs only.
o Receiver only Domains
Domains with only receivers get data without globally advertising
group membership.
Note that MSDP may be used with protocols other than PIM-SM, but such
usage is not specified in this memo.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119].
2. Overview
MSDP-speaking routers in a PIM-SM domain have a MSDP peering
relationship with MSDP peers in another domain. The peering
relationship is made up of a TCP connection in which control
information is exchanged. Each domain has one or more connections to
this virtual topology.
The purpose of this topology is to allow domains to discover
multicast sources from other domains. If the multicast sources are
of interest to a domain which has receivers, the normal source-tree
building mechanism in PIM-SM will be used to deliver multicast data
over an inter-domain distribution tree.
3. Procedure
When an RP in a PIM-SM domain first learns of a new sender, e.g., via
PIM register messages, it constructs a "Source-Active" (SA) message
and sends it to its MSDP peers. All RPs, which intend to originate
or receive SA messages, must establish MSDP peering with other RPs,
either directly or via an intermediate MSDP peer. The SA message
contains the following fields:
o Source address of the data source.
o Group address the data source sends to.
o IP address of the RP.
Note that an RP that isn’t a DR on a shared network SHOULD NOT
originate SA’s for directly connected sources on that shared network;
it should only originate in response to receiving Register messages
from the DR.
Each MSDP peer receives and forwards the message away from the RP
address in a "peer-RPF flooding" fashion. The notion of peer-RPF
flooding is with respect to forwarding SA messages. The Multicast
RPF Routing Information Base (MRIB) is examined to determine which
peer towards the originating RP of the SA message is selected. Such
a peer is called an "RPF peer". See section 13 for the details of
peer-RPF forwarding.
If the MSDP peer receives the SA from a non-RPF peer towards the
originating RP, it will drop the message. Otherwise, it forwards the
message to all its MSDP peers (except the one from which it received
the SA message).
When an MSDP peer which is also an RP for its own domain receives a
new SA message, it determines if there are any group members within
the domain interested in any group described by an (Source, Group),
or (S,G) entry within the SA message. That is, the RP checks for a
(*,G) entry with a non-empty outgoing interface list; this implies
that some system in the domain is interested in the group. In this
case, the RP triggers a (S,G) join event towards the data source as
if a Join/Prune message was received addressed to the RP itself.
This sets up a branch of the source-tree to this domain. Subsequent
data packets arrive at the RP via this tree branch, and are forwarded
down the shared-tree inside the domain. If leaf routers choose to
join the source-tree they have the option to do so according to
existing PIM-SM conventions. Finally, if an RP in a domain receives
a PIM Join message for a new group G, the RP SHOULD trigger a (S,G)
join event for each active (S,G) for that group in its SA cache.
This procedure has been affectionately named flood-and-join because
if any RP is not interested in the group, they can ignore the SA
message. Otherwise, they join a distribution tree.
4. Caching
A MSDP speaker MUST cache SA messages. Caching allows pacing of MSDP
messages as well as reducing join latency for new receivers of a
group G at an originating RP which has existing MSDP (S,G) state. In
addition, caching greatly aids in diagnosis and debugging of various
problems.
An MSDP speaker must provide a mechanism to reduce the forwarding of
new SA’s. The SA-cache is used to reduce storms and performs this by
not forwarding SA’s unless they are in the cache or are new SA
packets that the MSDP speaker will cache for the first time. The
SA-cache also reduces storms by advertising from the cache at a
period of no more than twice per SA-Advertisement-Timer interval and
not less than 1 time per SA Advertisement period.
5. Timers
The main timers for MSDP are: SA-Advertisement-Timer, SA Cache Entry
timer, Peer Hold Timer, KeepAlive timer, and ConnectRetry timer.
Each is considered below.
5.1. SA-Advertisement-Timer
RPs which originate SA messages do so periodically as long as there
is data being sent by the source. There is one SA-Advertisement-
Timer covering the sources that an RP may advertise. [SA-
Advertisement-Period] MUST be 60 seconds. An RP MUST not send more
than one periodic SA message for a given (S,G) within an SA
Advertisement interval. Originating periodic SA messages is required
to keep announcements alive in caches. Finally, an originating RP
SHOULD trigger the transmission of an SA message as soon as it
receives data from an internal source for the first time. This
initial SA message may be in addition to the periodic sa-message
forwarded in that first 60 seconds for that (S,G).
5.2. SA-Advertisement-Timer Processing
An RP MUST spread the generation of periodic SA messages (i.e.,
messages advertising the active sources for which it is the RP) over
its reporting interval (i.e., SA-Advertisement-Period). An RP starts
the SA-Advertisement-Timer when the MSDP process is configured. When
the timer expires, an RP resets the timer to [SA-Advertisement-
Period] seconds, and begins the advertisement of its active sources.
Active sources are advertised in the following manner: An RP packs
its active sources into an SA message until the largest MSDP packet
that can be sent is built or there are no more sources, and then
sends the message. This process is repeated periodically within the
SA-Advertisement-Period in such a way that all of the RP’s sources
are advertised. Note that since MSDP is a periodic protocol, an
implementation SHOULD send all cached SA messages when a connection
is established. Finally, the timer is deleted when the MSDP process
is de-configured.
5.3. SA Cache Timeout (SA-State Timer)
Each entry in an SA Cache has an associated SA-State Timer. A
(S,G)-SA-State-Timer is started when an (S,G)-SA message is initially
received by an MSDP peer. The timer is reset to [SG-State-Period] if
another (S,G)-SA message is received before the (S,G)-SA-State Timer
expires. [SG-State-Period] MUST NOT be less than [SA-Advertisement-
Period] + [SA-Hold-Down-Period].
5.4. Peer Hold Timer
The Hold Timer is initialized to [HoldTime-Period] when the peer’s
transport connection is established, and is reset to [HoldTime-
Period] when any MSDP message is received. Finally, the timer is
deleted when the peer’s transport connection is closed. [HoldTime-
Period] MUST be at least three seconds. The recommended value for
[HoldTime-Period] is 75 seconds.
5.5. KeepAlive Timer
Once an MSDP transport connection is established, each side of the
connection sends a KeepAlive message and sets a KeepAlive timer. If
the KeepAlive timer expires, the local system sends a KeepAlive
message and restarts its KeepAlive timer.
The KeepAlive timer is set to [KeepAlive-Period] when the peer comes
up. The timer is reset to [KeepAlive-Period] each time an MSDP
message is sent to the peer, and reset when the timer expires.
Finally, the KeepAlive timer is deleted when the peer’s transport
connection is closed.
[KeepAlive-Period] MUST be less than [HoldTime-Period], and MUST be
at least one second. The recommended value for [KeepAlive-Period] is
60 seconds.
5.6. ConnectRetry Timer
The ConnectRetry timer is used by the MSDP peer with the lower IP
address to transition from INACTIVE to CONNECTING states. There is
one timer per peer, and the [ConnectRetry-Period] SHOULD be set to 30
seconds. The timer is initialized to [ConnectRetry-Period] when an
MSDP speaker attempts to actively open a TCP connection to its peer
(see section 15, event E2, action A2 ). When the timer expires, the
peer retries the connection and the timer is reset to [ConnectRetry-
Period]. It is deleted if either the connection transitions into
ESTABLISHED state or the peer is de-configured.
6. Intermediate MSDP Peers
Intermediate MSDP speakers do not originate periodic SA messages on
behalf of sources in other domains. In general, an RP MUST only
originate an SA for a source which would register to it, and ONLY RPs
may originate SA messages. Intermediate MSDP speakers MAY forward SA
messages received from other domains.
7. SA Filtering and Policy
As the number of (S,G) pairs increases in the Internet, an RP may
want to filter which sources it describes in SA messages. Also,
filtering may be used as a matter of policy which at the same time
can reduce state. MSDP peers in transit domains should not filter SA
messages or the flood-and-join model can not guarantee that sources
will be known throughout the Internet (i.e., SA filtering by transit
domains may cause undesired lack of connectivity). In general,
policy should be expressed using MBGP [RFC2858]. This will cause
MSDP messages to flow in the desired direction and peer-RPF fail
otherwise. An exception occurs at an administrative scope [RFC2365]
boundary. In particular, a SA message for a (S,G) MUST NOT be sent
to peers which are on the other side of an administrative scope
boundary for G.
8. Encapsulated Data Packets
The RP MAY encapsulate multicast data from the source. An interested
RP may decapsulate the packet, which SHOULD be forwarded as if a PIM
register encapsulated packet was received. That is, if packets are
already arriving over the interface toward the source, then the
packet is dropped. Otherwise, if the outgoing interface list is
non-null, the packet is forwarded appropriately. Note that when
doing data encapsulation, an implementation MUST bound the time
during which packets are encapsulated.
This allows for small bursts to be received before the multicast tree
is built back toward the source’s domain. For example, an
implementation SHOULD encapsulate at least the first packet to
provide service to bursty sources.
9. Other Scenarios
MSDP is not limited to deployment across different routing domains.
It can be used within a routing domain when it is desired to deploy
multiple RPs for the same group ranges such as with Anycast RP’s. As
long as all RPs have a interconnected MSDP topology, each can learn
about active sources as well as RPs in other domains.
10. MSDP Peer-RPF Forwarding
The MSDP Peer-RPF Forwarding rules are used for forwarding SA
messages throughout an MSDP enabled internet. Unlike the RPF check
used when forwarding data packets, which generally compares the
packet’s source address against the interface upon which the packet
was received, the Peer-RPF check compares the RP address carried in
the SA message against the MSDP peer from which the message was
received.
10.1. Definitions
The following definitions are used in the description of the Peer-RPF
Forwarding Rules:
10.1.1. Multicast RPF Routing Information Base
The Multicast RPF Routing Information Base (MRIB) is the multicast
topology table. It is typically derived from the unicast routing
table or from other routing protocols such as multi-protocol BGP
[RFC2858].
10.1.2. Peer-RPF Route
The Peer-RPF route is the route that the MRIB chooses for a given
address. The Peer-RPF route for a SA’s originating RP is used to
select the peer from which the SA is accepted.
10.1.3. Peer-RPF Forwarding Rules
An SA message originated by R and received by X from N is accepted if
N is the peer-RPF neighbor for X, and is discarded otherwise.
MPP(R,N) MP(N,X)
R ---------....-------> N ------------------> X
SA(S,G,R) SA(S,G,R)
MP(N,X) is an MSDP peering between N and X. MPP(R,N) is an MSDP
peering path (zero or more MSDP peers) between R and N, e.g.,
MPP(R,N) = MP(R, A) + MP(A, B) + MP(B, N). SA(S,G,R) is an SA
message for source S on group G originated by an RP R.
The peer-RPF neighbor N is chosen deterministically, using the first
of the following rules that matches. In particular, N is the RPF
neighbor of X with respect to R if
(i). N == R (X has an MSDP peering with R).
(ii). N is the eBGP NEXT_HOP of the Peer-RPF route for R.
(iii). The Peer-RPF route for R is learned through a distance-vector
or path-vector routing protocol (e.g., BGP, RIP, DVMRP) and N
is the neighbor that advertised the Peer-RPF route for R
(e.g., N is the iBGP advertiser of the route for R), or N is
the IGP next hop for R if the route for R is learned via a
link-state protocol (e.g., OSPF [RFC2328] or IS-IS
[RFC1142]).
(iv). N resides in the closest AS in the best path towards R. If
multiple MSDP peers reside in the closest AS, the peer with
the highest IP address is the rpf-peer.
(v). N is configured as the static RPF-peer for R.
MSDP peers, which are NOT in state ESTABLISHED (i.e., down peers),
are not eligible for peer RPF consideration.
10.2. MSDP mesh-group semantics
An MSDP mesh-group is a operational mechanism for reducing SA
flooding, typically in an intra-domain setting. In particular, when
some subset of a domain’s MSDP speakers are fully meshed, they can be
configured into a mesh-group.
Note that mesh-groups assume that a member doesn’t have to forward an
SA to other members of the mesh-group because the originator will
forward to all members. To be able for the originator to forward to
all members (and to have each member also be a potential originator),
the mesh-group must be a full mesh of MSDP peering among all members.
The semantics of the mesh-group are as follows:
(i). If a member R of a mesh-group M receives a SA message from an
MSDP peer that is also a member of mesh-group M, R accepts
the SA message and forwards it to all of its peers that are
not part of mesh-group M. R MUST NOT forward the SA message
to other members of mesh-group M.
(ii). If a member R of a mesh-group M receives an SA message from
an MSDP peer that is not a member of mesh-group M, and the SA
message passes the peer-RPF check, then R forwards the SA
message to all members of mesh-group M and to any other msdp
peers.
11. MSDP Connection State Machine
MSDP uses TCP as its transport protocol. In a peering relationship,
one MSDP peer listens for new TCP connections on the well-known port
639. The other side makes an active connect to this port. The peer
with the higher IP address will listen. This connection
establishment algorithm avoids call collision. Therefore, there is
no need for a call collision procedure. It should be noted, however,
that the disadvantage of this approach is that the startup time
depends completely upon the active side and its connect retry timer;
the passive side cannot cause the connection to be established.
An MSDP peer starts in the DISABLED state. MSDP peers establish
peering sessions according to the following state machine:
--------------->+----------+
/ | DISABLED |<----------
| ------>+----------+ \
| / |E1->A1 |
| | | |
| | V |E7->A7
| | +----------+ E3->A3 +--------+
| | | INACTIVE |------->| LISTEN |
| | +----------+ +--------+
| | E2->A2| ^ |E5->A5
| | | | |
| |E7->A6 V |E6 |
| \ +------------+ |
| ------| CONNECTING | |
| +------------+ |