Multicast Fundamentals: IGMP and PIM Explained

Sending the same video stream individually to a thousand viewers would waste enormous bandwidth, and multicast solves this by delivering a single stream efficiently to exactly the devices that actually want it. This article explains how multicast addressing differs from unicast and broadcast, covers IGMP as the protocol hosts use to join multicast groups, and walks through how PIM builds the distribution trees that carry multicast traffic efficiently through a network.

IP MulticastIGMPPIM Sparse Mode

~6 min read · Updated Sep 11, 2026

Why Sending Data to Many Receivers Needs a Different Approach

The unicast addressing discussed throughout most of this series delivers traffic to exactly one destination. Broadcast, discussed earlier in this series, delivers to every device on a local segment, regardless of whether they actually want the traffic. Neither approach scales well for a scenario like a live video stream watched by a thousand interested users spread across many different subnets: unicast would require sending a thousand separate copies of the same stream, while broadcast would flood the traffic to devices that have no interest in it at all, and could not cross subnet boundaries in the first place.

Multicast: One Stream, Many Interested Receivers

IP Multicast solves this by delivering a single stream of traffic to a special multicast group address, and having the network itself replicate that traffic only at the points where it actually needs to branch toward multiple interested receivers.

Multicast address range: 224.0.0.0 to 239.255.255.255
  (Class D, discussed earlier in this series
  regarding historical address classes)

A source sends a single stream to a group address,
such as 239.1.1.1, and only devices that have
explicitly asked to receive traffic for that
specific group actually get a copy of it

How a Host Joins a Multicast Group: IGMP

IGMP (Internet Group Management Protocol) is how a host informs its local router that it wants to receive traffic for a specific multicast group.

IGMP message types:

Membership Report: sent by a host to join a
  specific multicast group

Query: sent periodically by the router to check
  whether any host on the segment still wants
  traffic for a given group

Leave Group: sent by a host to explicitly leave
  a group (in IGMPv2 and later), allowing the
  router to stop forwarding that group's traffic
  to the segment more quickly than waiting for
  a query timeout

Router(config)# interface gigabitethernet 0/1
Router(config-if)# ip igmp version 2

Router# show ip igmp groups

IGMP Connected Group Membership
Group Address    Interface         Uptime    Expires
239.1.1.1         Gi0/1             00:15:22  00:02:30

The router uses this group membership information to decide whether it needs to forward a given multicast stream onto a specific segment at all — if no host on that segment has joined the group, the router does not waste bandwidth sending traffic that nothing is listening for.

Building Distribution Trees: PIM

While IGMP handles the "last hop" between a router and its directly connected hosts, PIM (Protocol Independent Multicast) handles the harder problem of building an efficient distribution path across the routers between the multicast source and every interested receiver, potentially spanning many hops and multiple subnets.

The term "Protocol Independent" refers to PIM
relying on the existing unicast routing table,
discussed earlier in this series, built by
whatever unicast routing protocol (OSPF, EIGRP,
BGP) is already running -- PIM does not maintain
its own separate topology database, instead
using the unicast routing table to determine
the correct direction back toward a multicast
source

PIM Sparse Mode: The Modern Standard

PIM Sparse Mode (PIM-SM) is the dominant PIM mode in modern networks, built around the assumption that multicast receivers are relatively sparse and spread out, so traffic should only be sent to network segments that have explicitly requested it — the polar opposite assumption of the largely obsolete PIM Dense Mode, which initially floods traffic everywhere and then prunes back branches with no interested receivers.

Key PIM-SM concept: the Rendezvous Point (RP)

The RP is a specifically designated router that
acts as a known meeting point: multicast sources
initially register with the RP, and receivers
initially join toward the RP rather than needing
to know a source's location directly

Router(config)# ip pim rp-address 10.0.0.1

-- Configures which router acts as the RP for
-- the multicast domain -- every PIM-enabled
-- router must be configured with the same
-- RP address to function correctly together

The Shared Tree Versus the Shortest-Path Tree

PIM-SM initially builds a Shared Tree rooted at the RP, which lets receivers join a group without needing to know where the source actually is. Once traffic begins flowing, PIM-SM can optionally switch to a more efficient Shortest-Path Tree (SPT) that bypasses the RP entirely, routing directly between the source and each receiver along the actual shortest path.

Traffic flow evolution in PIM-SM:

Initial state: Source → RP → Shared Tree → Receivers
  (traffic takes a potentially longer path through
   the RP, but this works immediately without
   receivers needing prior knowledge of the source)

After SPT switchover: Source → Shortest-Path Tree
  → Receivers directly (bypassing the RP entirely,
  once the last-hop router has learned the actual
  source address and determined a more direct
  path exists)

This two-phase approach elegantly balances two competing needs: receivers can join a multicast group immediately without any prior knowledge of where the source is located, while long-term traffic flow still converges toward the genuinely most efficient path once that information becomes available.

Verifying Multicast Routing State

Router# show ip mroute

(*, 239.1.1.1), 00:20:15/00:02:45, RP 10.0.0.1, flags: S
  Incoming interface: GigabitEthernet0/0
  Outgoing interface list:
    GigabitEthernet0/1, Forward/Sparse, 00:15:22/00:02:30

(192.168.1.10, 239.1.1.1), 00:05:03/00:03:15, flags: T
  Incoming interface: GigabitEthernet0/2
  Outgoing interface list:
    GigabitEthernet0/1, Forward/Sparse, 00:05:03/00:03:15

The multicast routing table shows two distinct entry types: a (*, G) entry (any source, specific group) representing the shared tree rooted at the RP, and an (S, G) entry (specific source, specific group) representing the shortest-path tree once switchover has occurred — reading this output directly reveals which phase of the two-phase process described above a given multicast flow is currently in.

Why Multicast Remains Essential for Specific Applications

While the internet at large has increasingly moved toward unicast-based content delivery for many streaming applications, IP multicast remains essential in specific environments: financial market data distribution, IPTV within managed service provider networks, and certain enterprise applications requiring efficient one-to-many communication within a controlled network. Understanding IGMP's role at the network edge and PIM's role building efficient distribution trees across the network core provides the foundation for correctly deploying and troubleshooting any of these multicast-dependent applications.

Written & researched by Dr. Shahin Siami

Related Articles

SDN and SD-WAN Fundamentals: Separating the Control Plane from the Data Plane

Traditional networking, covered throughout most of this series, distributes intelligence across every individual device, each making its own independent forwarding decisions. Software-Defined Networking fundamentally changes this by centralizing that intelligence, and this article explains the control plane and data plane separation underlying SDN, covers how SD-WAN applies these principles specifically to wide area network connectivity, and explains the practical benefits this architectural shift provides.

Continue

Enterprise Network Architecture: The Three-Tier Design Model

Every technology covered so far in this series -- VLANs, routing protocols, redundancy protocols -- needs an overall architectural framework to be deployed coherently rather than as an ad hoc collection of features. This article explains the classic three-tier hierarchical design model, covers the distinct role each layer plays, explains the simplified two-tier collapsed core alternative, and discusses how these models extend into modern data center design.

Continue

First Hop Redundancy Protocols: HSRP, VRRP, and GLBP Explained

Every host on a network relies on a single default gateway, and that gateway becoming a single point of failure would undermine the redundancy carefully built everywhere else in the network. This article explains why first hop redundancy matters, walks through HSRP's active/standby model, compares it against the open-standard VRRP, and covers GLBP's added ability to load-balance traffic across multiple routers simultaneously.

Continue

Route Redistribution: Exchanging Routes Between Different Routing Protocols

Real enterprise networks often run multiple routing protocols simultaneously, whether due to mergers, legacy equipment, or vendor requirements, and these protocols do not automatically share routes with each other. This article explains why redistribution becomes necessary, covers the critical metric mismatch problem between protocols, walks through configuring redistribution between OSPF and EIGRP, and covers the routing loop risks that make careful redistribution design essential.

Continue

BGP Fundamentals: The Protocol That Runs the Internet

Every interior routing protocol covered so far in this series operates within a single organization's network, but connecting separate organizations together across the internet requires an entirely different protocol built around policy rather than pure shortest-path calculation. This article explains what makes BGP a path-vector protocol, covers the distinction between eBGP and iBGP, walks through essential path attributes used for path selection, and covers basic BGP configuration and verification.

Continue

Multi-Area OSPF: Scaling with Areas, LSA Types, and Route Summarization

A single-area OSPF design, discussed earlier in this series, does not scale to large networks, since every router must process the full topology database of every other router. This article explains why OSPF areas exist, covers the different Link-State Advertisement types that carry information between areas, explains the role of Area Border Routers, and walks through configuring route summarization to keep large multi-area networks efficient.

Continue