Network Working Group D. Farinacci Request for Comments: 4610 Y. Cai Category: Standards Track Cisco Systems August 2006
Anycast-RP Using Protocol Independent Multicast (PIM)
Status of This Memo
This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2006).
Abstract
This specification allows Anycast-RP (Rendezvous Point) to be used inside a domain that runs Protocol Independent Multicast (PIM) only. Other multicast protocols (such as Multicast Source Discovery Protocol (MSDP), which has been used traditionally to solve this problem) are not required to support Anycast-RP.
Anycast-RP as described in [I1] is a mechanism that ISP-based backbones have used to get fast convergence when a PIM Rendezvous Point (RP) router fails. To allow receivers and sources to Rendezvous to the closest RP, the packets from a source need to get to all RPs to find joined receivers.
This notion of receivers finding sources is the fundamental problem of source discovery that MSDP was intended to solve. However, if one would like to retain the Anycast-RP benefits from [I1] with less protocol machinery, removing MSDP from the solution space is an option.
This memo extends the Register mechanism in PIM so Anycast-RP functionality can be retained without using MSDP.
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 [N2].
o A unicast IP address is chosen to use as the RP address. This address is statically configured, or distributed using a dynamic protocol, to all PIM routers throughout the domain.
o A set of routers in the domain is chosen to act as RPs for this RP address. These routers are called the Anycast-RP set.
o Each router in the Anycast-RP set is configured with a loopback interface using the RP address.
o Each router in the Anycast-RP set also needs a separate IP address, to be used for communication between the RPs.
o The RP address, or a prefix that covers the RP address, is injected into the unicast routing system inside of the domain.
o Each router in the Anycast-RP set is configured with the addresses of all other routers in the Anycast-RP set. This must be consistently configured in all RPs in the set.
The following diagram illustrates a domain using 3 RPs where receivers are joining to the closest RP according to where unicast routing metrics take them and 2 sources sending packets to their respective RPs.
The rules described in this section do not override the rules in [N1]. They are intended to blend with the rules in [N1]. If there is any question on the interpretation, precedent is given to [N1].
S1-----RP1 RP2 RP3------S3 / \ | / \ | R1 R1' R2
Farinacci & Cai Standards Track [Page 2]
RFC 4610 Anycast-RP using PIM August 2006
Assume the above scenario is completely connected where R1, R1', and R2 are receivers for a group, and S1 and S3 send to that group. Assume RP1, RP2, and RP3 are all assigned the same IP address, which is used as the Anycast-RP address (let's say the IP address is RPA).
Note, the address used for the RP address in the domain (the Anycast-RP address) needs to be different than the addresses used by the Anycast-RP routers to communicate with each other.
The following procedure is used when S1 starts sourcing traffic:
o S1 sends a multicast packet.
o The designated router (DR) directly attached to S1 will form a PIM Register message to send to the Anycast-RP address (RPA). The unicast routing system will deliver the PIM Register message to the nearest RP, in this case RP1.
o RP1 will receive the PIM Register message, decapsulate it, and send the packet down the shared-tree to get the packet to receivers R1 and R1'.
o RP1 is configured with RP2 and RP3's IP address. Since the Register message did not come from one of the RPs in the anycast-RP set, RP1 assumes the packet came from a DR. If the Register is not addressed to the Anycast-RP address, an error has occurred and it should be rate-limited logged.
o RP1 will then send a copy of the Register message from S1's DR to both RP2 and RP3. RP1 will use its own IP address as the source address for the PIM Register message.
o RP1 MAY join back to the source-tree by triggering a (S1,G) Join message toward S1. However, RP1 MUST create (S1,G) state.
o RP1 sends a Register-Stop back to the DR. If, for some reason, the Register messages to RP2 and RP3 are lost, then when the Register suppression timer expires in the DR, it will resend Registers to allow another chance for all RPs in the Anycast-RP set to obtain the (S,G) state.
o RP2 receives the Register message from RP1, decapsulates it, and also sends the packet down the shared-tree to get the packet to receiver R2.
o RP2 sends a Register-Stop back to RP1. RP2 MAY wait to send the Register-Stop if it decides to join the source-tree. RP2 should wait until it has received data from the source on the source-tree
Farinacci & Cai Standards Track [Page 3]
RFC 4610 Anycast-RP using PIM August 2006
before sending the Register-Stop. If RP2 decides to wait, the Register-Stop will be sent when the next Register is received. If RP2 decides not to wait, the Register-Stop is sent now.
o RP2 MAY join back to the source-tree by triggering a (S1,G) Join message toward S1. However, RP2 MUST create (S1,G) state.
o RP3 receives the Register message from RP1, decapsulates it, but since there are no receivers joined for the group, it can discard the packet.
o RP3 sends a Register-Stop back to RP1.
o RP3 creates (S1,G) state so when a receiver joins after S1 starts sending, RP3 can join quickly to the source-tree for S1.
o RP1 processes the Register-Stop from each of RP2 and RP3. There is no specific action taken when processing Register-Stop messages.
The procedure for S3 sending follows the same as above but it is RP3 that sends a copy of the Register originated by S3's DR to RP1 and RP2. Therefore, this example shows how sources anywhere in the domain, associated with different RPs, can reach all receivers, also associated with different RPs, in the same domain.
4. Observations and Guidelines about This Proposal
o An RP will send a copy of a Register only if the Register is received from an IP address not in the Anycast-RP list (i.e., the Register came from a DR and not another RP). An implementation MUST safeguard against inconsistently configured Anycast-RP sets in each RP by copying the Time to Live (TTL) from a Register message to the Register messages it copies and sends to other RPs.
o Each DR that PIM registers for a source will send the message to the Anycast-RP address (which results in the packet getting to the closest physical RP). Therefore, there are no changes to the DR logic.
o Packets flow to all receivers no matter what RP they have joined to.
o The source gets Registered to a single RP by the DR. It's the responsibility of the RP that receives the PIM Register messages from the DR (the closest RP to the DR based on routing metrics) to get the packet to all other RPs in the Anycast-RP set.
Farinacci & Cai Standards Track [Page 4]
RFC 4610 Anycast-RP using PIM August 2006
o Logic is changed only in the RPs. The logic change is for sending copies of Register messages. Register-Stop processing is unchanged. However, an implementation MAY suppress sending Register-Stop messages in response to a Register received from an RP.
o The rate-limiting of Register and Register-Stop messages are done end-to-end. That is from DR -> RP1 -> {RP2 and RP3}. There is no need for specific rate-limiting logic between the RPs.
o When topology changes occur, the existing source-tree adjusts as it does today according to [N1]. The existing shared-trees, as well, adjust as they do today according to [N1].
o Physical RP changes are as fast as unicast route convergence, retaining the benefit of [I1].
o An RP that doesn't support this specification can be mixed with RPs that do support this specification. However, the non-supporter RP should not have sources registering to it, but may have receivers joined to it.
o If Null Registers are sent (Registers with an IP header and no IP payload), they MUST be replicated to all of the RPs in the Anycast-RP set so that source state remains alive for active sources.
o The number of RPs in the Anycast-RP set should remain small so the amount of non-native replication is kept to a minimum.
o Since the RP, who receives a Register from the DR, will send copies of the Register to the other RPs at the same time it sends a Register-Stop to the DR, there could be packet loss and lost state in the other RPs until the time the DR sends Register messages again.
5. Interaction with MSDP Running in an Anycast-PIM Router
The objective of this Anycast-PIM proposal is to remove the dependence on using MSDP. This can be achieved by removing MSDP peering between the Anycast-RPs. However, to advertise internal sources to routers outside of a PIM routing domain and to learn external sources from other routing domains, MSDP may still be required.
In this capacity, when there are internal sources that need to be advertised externally, an Anycast-RP that receives a Register message, either from a DR or an Anycast-RP, should process it as described in this specification as well as how to process a Register message as described in [N1]. That means a Source-Active (SA) for the same internal source could be originated by multiple Anycast-RPs doing the MSDP peering. There is nothing inherently wrong with this other than that the source is being advertised into the MSDP infrastructure from multiple places from the source domain. However, if this is not desirable, configuration of one or more (rather than all) Anycast-RP MSDP routers would allow only those routers to originate SAs for the internal source. And in some situations, there is a good possibility not all Anycast-RPs in the set will have MSDP peering sessions so this issue can be mitigated to a certain extent.
From an Anycast-RP perspective, a source should be considered internal to a domain when it is discovered by an Anycast-RP through a received Register message, regardless of whether the Register message was sent by a DR, another Anycast-RP member, or the router itself.
For learning sources external to a domain, the MSDP SA messages could arrive at multiple MSDP-peering Anycast-RPs. The rules for processing an SA, according to [I1], should be followed. That is, if G is joined in the domain, an (S,G) join is sent towards the source. And if data accompanies the SA, each Anycast-PIM RP doing MSDP peering will forward the data down each of its respective shared- trees.
The above assumes each Anycast-RP has external MSDP peering connections. If this is not the case, the Anycast-PIM routers with the MSDP peering connections would follow the same procedure as if a Data-Register or Null-Register was received from either a DR or another Anycast-RP. That is, they would send Registers to the other members of the Anycast-RP set.
If there is a mix of Anycast-RPs that do and do not have external MSDP peering connections, then the ones that do must be configured with the set that do not. So Register messages are sent only to the members of the Anycast-RP set that do not have external MSDP peering connections.
The amount of Register traffic generated by this MSDP-peering RP would be equal to the number of active sources external to the domain. The Source-Active state would have to be conveyed to all other RPs in the Anycast-RP set since the MSDP-peering RP would not know about the group membership associated with the other RPs. To
Farinacci & Cai Standards Track [Page 6]
RFC 4610 Anycast-RP using PIM August 2006
avoid this periodic control traffic, it is recommended that all Anycast-RPs be configured with external MSDP peering sessions so no RP in the Anycast-RP set will have to originate Register messages on behalf of external sources.
Within a routing domain, it is recommended that an Anycast-RP set defined in this specification should not be mixed with MSDP peering among the members. In some cases, the source discovery will work but it may not be obvious to the implementations which sources are local to the domain and which are not. This may affect external MSDP advertisement of internal sources.
Having said that, this document makes no attempt to connect MSDP peering domains together by using Anycast-PIM inside a transit domain.
This section describes the security consideration for Register and Register-Stop messages between Anycast-RPs. For PIM messages between DR and RP, please see [N1].
An attacker may forge a Register message using one of the addresses in the Anycast-RP list in order to achieve one or more of the following effects:
1. Overwhelm the target RP in a denial-of-service (DoS) attack 2. Inject unauthorized data to receivers served by the RP 3. Inject unauthorized data and create bogus SA entries in other PIM domains if the target RP has external MSDP peerings
An attacker may also forge a Register-Stop message using one of the addresses in the Anycast-RP list. However, besides denial-of- service, the effect of such an attack is limited because an RP usually ignores Register-Stop messages.
The DoS attack using forged Register or Register-Stop messages cannot be prevented. But the RP can still be protected. For example, the RP can rate-limit incoming messages. It can also choose to refuse to process any Register-Stop messages. The actual protection mechanism is implementation specific.
Farinacci & Cai Standards Track [Page 7]
RFC 4610 Anycast-RP using PIM August 2006
The distribution of unauthorized data and bogus Register messages can be prevented using the method described in section 6.3.2 of [N1]. When RP1 sends a copy of a register to RP2, RP1 acts as [N1] describes the DR and RP2 acts as [N1] describes the RP.
As described in [N1], an RP can be configured using a unique SA and Security Parameter Index (SPI) for traffic (Registers or Register- Stops) to each member of Anycast-RPs in the list, but this results in a key management problem; therefore, it may be preferable in PIM domains where all Rendezvous Points are under a single administrative control to use the same authentication algorithm parameters (including the key) for all Registered packets in a domain.
The authors prototyped this document in the cisco IOS and Procket implementations, respectively.
The authors would like to thank John Zwiebel for doing interoperability testing of the two prototype implementations.
The authors would like to thank Thomas Morin from France Telecom for having an extensive discussion on Multicast the Registers to an SSM- based full mesh among the Anycast-RP set. This idea may come in a subsequent document.
And finally, the authors would like to thank the following for their comments on earlier drafts:
Greg Shepherd (Procket Networks (now Cisco Systems)) Lenny Giuliano (Juniper Networks) Prashant Jhingran (Huawei Technologies) Pekka Savola (CSC/FUNET) Bill Fenner (AT&T) James Lingard (Data Connection) Amit Shukla (Juniper Networks) Tom Pusateri (Juniper Networks)
[I1] Kim, D., Meyer, D., Kilmer, H., and D. Farinacci, "Anycast Rendevous Point (RP) mechanism using Protocol Independent Multicast (PIM) and Multicast Source Discovery Protocol (MSDP)", RFC 3446, January 2003.
<anycast-rp-addr> describes the Anycast-RP set for the RP that is assigned to the group range. This IP address is the address that first-hop and last-hop PIM routers use to register and join to.
<rp-addr> describes the IP address where Register messages copies are sent to. This IP address is any address assigned to the RP router not including the <anycast-rp-addr>.
Example:
From the illustration above, the configuration commands would be:
ip pim anycast-rp RPA RP1 ip pim anycast-rp RPA RP2 ip pim anycast-rp RPA RP3
Comment:
It may be useful to include the local router IP address in the command set so the above lines can be cut-and-pasted or scripted into all the RPs in the Anycast-RP set.
But the implementation would have to be aware of its own address and not inadvertently send a Register to itself.
Farinacci & Cai Standards Track [Page 10]
RFC 4610 Anycast-RP using PIM August 2006
Authors' Addresses
Dino Farinacci Cisco Systems
EMail: dino@cisco.com
Yiqun Cai Cisco Systems
EMail: ycai@cisco.com
Farinacci & Cai Standards Track [Page 11]
RFC 4610 Anycast-RP using PIM August 2006
Full Copyright Statement
Copyright (C) The Internet Society (2006).
This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights.
This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Intellectual Property
The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org.
Acknowledgement
Funding for the RFC Editor function is provided by the IETF Administrative Support Activity (IASA).