The Rendezvous Point (RP) is a fundamental piece in all variants of the Any Source Multicast (ASM) distribution model. In any network design with ASM involved, high availability of the RP should be taken into account in order to increase the overall design availability and scalability.
For each multicast Group in a multicast domain, the configuration of the Active RP should be coherent along the domain to ensure that there is only one RP Active simultaneously on that domain. The Active RP can be elected by using an automatic information distribution mechanism, like Bootstrap Router (BSR), or by manually configuring the RP in every network device (Static RP). Static RP configuration does not provide RP redundancy by default. Anycast RP and/or Phantom RP are two RP redundancy mechanisms that can be used to provide redundancy to Static RP configuration.
In large networks where multicast Groups can have a high number of sources and receivers, the network designer can define several RPs to provide high availability to the multicast design. Load sharing configurations allows several RPs to share the load of Source Registration. For redundancy purposes, more than one RP may be configured to improve the availability of the multicast solution.
RP design considerations overview
There are two main RP redundancy mechanisms applicable to the ASM multicast distribution model: Anycast RP and Phantom RP. Anycast RP is a redundancy and load balancing mechanism applicable to the PIM-SM distribution model and Phantom RP is designed to provide a backup scenario to PIM-BiDir designs. Both solutions provide load balancing and redundancy to the Static RP configuration method.
RP redundancy mechanism
Before getting into details about the backup solutions for the RP is important to know the role and functionality of the RP in the two variants of the ASM distribution model.
There are some differences in the functionality for the RP in the Control Plane depending on the PIM variant and its associated use case. When using PIM-SM, the RP acts as a meeting point between sources and receivers. There is load associated to Source Registering and it makes sense to think about load sharing in the control plane. On the contrary, in PIM-BiDir the RP doesn’t play any specific role in the control plane.
In the Data Plane there are also important differences. When using PIM-SM, multicast traffic flows at first from the Source to the RP and from there using a Shared Tree down to the receivers. If Shortest Path Tree (SPT) switchover is ON, then the data would flow directly from Sources to Receivers via a SPT. The RP failure would have a direct impact only on new Sources or Receivers.
In PIM-BiDir, there are only Bidirectional Shared Trees, and all the packets move up and down the shared tree in a bidirectional way. The RP failure has a direct impact on all sources sharing the same shared tree. In this case, the RP is always in the data path, and the RP position is more relevant than in the PIM-SM case.
Anycast RP is a mechanism used to define redundant load-balanced RP devices. The applicability of Anycast RP is limited to the ASM distribution model in combination with the PIM-SM variant.
Two or more RP can be configured to share the load for source registration and act as hot backup one of each other. Anycast RP can be used with Static RP configuration to provide failover and redundancy within a PIM-SM domain.
Working mechanism of Anycast RP for ASM with PIM-SM:
- PIM register -> sent to the closest RP (load balancing for source registering).
- PIM join/prune -> sent in the direction of the closest RP as determined by the unicast routing table (traffic load balancing if SPT switchover is OFF).
If one of the RPs goes down, unicast routing ensures that these messages will be sent in the direction of the next-closest RP. This functionality has 2 possible implementations:
- Using MSDP (RFC 3446); limited to IPv4 only
- Using PIM (RFC 4610); IPv6 compatible
In any of the two cases, the source distribution protocol will share the source information with any of the members of the RP set for the PIM domain.
MSDP is the protocol used to share information about active sources. MSDP applicability is limited to IPv4 deployments as there is not interoperable with IPv6.
Multicast routers use the closest RP based on the unicast routing table (load sharing) and in case of failure, Receivers join… and Sources register… to the new-closest RP (hot-backup).
PIM Anycast RP allows Anycast RP functionality in PIM-SM domains that run only PIM or IPv6 deployments. PIM Anycast RP extends the register mechanisms in PIM so that Anycast RP functionality can be achieved without using MSDP.
In case of failure, the RP convergence time depends on the IGP convergence time (RP convergence = IGP convergence time).
The RP redundancy mechanism for PIM-BiDir is called Phantom RP. In PIM-BiDir, the RP does not have an actual protocol function, it doesn’t act as a meeting point neither has source registering functionality associated, the RP is only a reference on the network. The RP acts as a routing vector pointing to a place where all the traffic converges.
The RP can be configured as an address not assigned to any particular device called Phantom RP. The RP address doesn’t need to reside on any physical router interface. Can be just an IP address in a subnet that will be advertised with two different length masks by routing devices on the network via the IGP.
The RP is called Virtual RP because doesn’t need to be a physical router. The RP address is just an indication towards the root of the Shared Tree.
In case of failure, the RP convergence time depends on the IGP convergence time and the DF election delay (RP convergence = IGP convergence time + DF election delay).
Depending on the model used, the RP associated functionality in the control and data planes varies, and so it does the RP backup mechanism.
Anycast RP could be used to provide hot backup and load sharing functionality to PIM-SM multicast domains. Phantom RP should be considered to provide a backup mechanism for PIM-BiDir.
Doesn’t play any role
Not very relevant
In the control plane, the role of the RP in PIM-SM is to act as a meeting point between Sources and Receivers. RP failure scenario has a limited impact compared with PIM-BiDir, because if SPT switchover is ON, the multicast data traffic would continue flowing via SPT from Sources to Receivers, affecting only to new source discovery.
In PIM-BiDir based designs, the location of the RP is very relevant because it’s always located in the data plane between Sources and Receivers. The RP failure is critical and a backup should be always considered because has a direct impact on all data Sources.
I hope you find it useful.