Protocol-Independent Multicast - Dense Mode

Description

Protocol-Independent Multicast Dense Mode (PIM-DM) is one of the two primary operational modes of the PIM protocol suite, the other being PIM Sparse Mode (PIM-SM). PIM-DM operates under the assumption that multicast group members are densely distributed throughout the network, meaning that most subnets will have at least one receiver for any given multicast group. Its fundamental operation follows a 'flood-and-prune' or 'push' model. When a multicast source begins sending traffic, PIM-DM initially floods this traffic out of all PIM-enabled interfaces (except the one it arrived on, based on RPF check). This creates a shortest-path tree rooted at the source.

The protocol mechanism involves periodic flooding followed by explicit pruning. When a router receives multicast data on an interface and has no downstream receivers (as determined by lacking IGMP/MLD membership reports or downstream PIM joins), it sends a PIM Prune message upstream towards the source. This prunes that branch off the distribution tree. These pruned states have a lifetime; after expiry, the flooding resumes, and the prune process may repeat if receivers are still absent. PIM-DM also uses a Graft mechanism to quickly re-join a previously pruned tree when a new receiver appears, sending a Graft message directly to the upstream router to re-activate the branch without waiting for the next flood cycle.

Key architectural elements include the use of the underlying unicast routing table for RPF checks and the maintenance of source-specific (S,G) state in every router along the potential path. Unlike PIM-SM, PIM-DM does not use a Rendezvous Point (RP). It builds source trees immediately. State is refreshed by the periodic flooding of data and associated control messages.

Within 3GPP specifications, PIM-DM is explicitly mentioned in TS 29.561 (5G core exposure) as a supported multicast routing protocol. Its role in mobile networks is typically for specific deployment scenarios within the service provider's IP transport network or for certain types of multicast services where the dense mode assumption holds, such as delivering a common channel to all nodes in a confined network segment. It provides a simple, low-latency method for multicast distribution without the need for pre-configuration of RPs.

Purpose & Motivation

PIM-DM was designed for network environments where multicast group members are densely populated, meaning receivers are present on a high percentage of subnets. The purpose is to provide a simple, automatic mechanism to build source-based multicast distribution trees with minimal configuration. It addresses the need for efficient many-to-many communication within a confined, receiver-dense network, such as a corporate LAN or a specific data center fabric.

The flood-and-prune model solves the problem of initial join latency. In PIM-SM, a receiver must wait for a join to propagate to an RP before receiving data. PIM-DM, by immediately flooding data, can deliver the first packets with minimal delay, which can be beneficial for certain applications. However, this comes at the cost of initial bandwidth waste on branches without receivers, which is later reclaimed via pruning. PIM-DM is therefore motivated by scenarios where bandwidth is plentiful relative to the number of subnets, and low initial latency is valued over optimal bandwidth efficiency from the start.

In the context of 3GPP's adoption, PIM-DM provides an alternative multicast routing protocol option for the transport networks that underpin mobile core functions. It offers network operators flexibility. For instance, within a tightly coupled data center hosting 5G core functions, PIM-DM could be used for efficient internal multicast distribution of operational data. Its specification in 3GPP documents ensures that network exposure functions (NEF) and other elements can be designed to interoperate with networks running PIM-DM, supporting a wider range of operator deployment choices.

Key Features

• Operates on a flood-and-prune (push) model for tree building.
• Assumes dense receiver distribution and initially floods data to all branches.
• Uses Prune messages to cut off branches without active receivers.
• Employs Graft messages for fast re-joining of pruned branches when receivers appear.
• Builds source-specific (S,G) shortest-path trees without requiring a Rendezvous Point (RP).
• Relies on periodic flooding to refresh state and rediscover new receivers.

Evolution Across Releases

Defining Specifications