Quick Summary
- Unicast is one-to-one (a single sender to a single receiver) and makes up the vast majority of internet traffic
- Broadcast is one-to-all on the local network — every host in the broadcast domain gets it, and routers never forward it
- Multicast is one-to-many: a single stream goes only to receivers that joined the group, making it far more efficient than many unicast copies
The Phone Call, the Announcement, and the Newsletter
Imagine you have some information to share, and three different ways to deliver it. A unicast is a private phone call — you dial one specific person, and only they hear what you say. A broadcast is grabbing the building intercom and shouting an announcement to everyone, whether they care about it or not. A multicast is a subscription newsletter: only the people who actually signed up receive a copy, and everyone else is left in peace.
Networks deliver packets in exactly these three ways. The destination of a packet decides which mode it is: a single device, every device on the local segment, or a hand-picked group of devices that have opted in.
Choosing the right mode matters because it controls who is bothered and how many copies get made. Unicast is precise but does not scale when you need to reach many receivers. Broadcast reaches everyone instantly but wastes effort on hosts that do not care. Multicast threads the needle — it reaches many interested receivers while sending just one stream that the network replicates only where it has to.
Visualizing the Three Modes
Here one sender talks to six hosts on a switch. The hosts marked with a star have subscribed to a multicast group. Watch how each mode reaches a different set of receivers:
Delivery Mode Visualizer
One sender, six hosts on the same subnet. Two hosts joined the multicast group (★). Pick a mode and watch who gets a copy — and how many copies the sender makes.
Note: "packets sent by source" reflects how many copies the sender itself must originate — the metric that makes multicast scale.
Choose a delivery mode above to begin.
Notice the counters: unicast and broadcast both originate a single packet at the source, but they reach very different numbers of hosts. The real efficiency story shows up when you need to reach many receivers: to serve 100 viewers, unicast forces the source to send 100 separate copies, while multicast sends one stream that the network replicates only where the paths to receivers split.
The Three Modes in Detail
Unicast: One-to-One
Unicast is a conversation between exactly two devices — one sender, one specific receiver, identified by that receiver's own IP and MAC address. The overwhelming majority of internet traffic is unicast: loading a web page, sending an email, downloading a file, or streaming a video to your single device are all unicast. It is precise and reliable, but it does not scale when the same content must reach many receivers, because the source must transmit an independent copy to each one.
Broadcast: One-to-All
Broadcast delivers a packet to every host on the local network or subnet — the whole
broadcast domain. In IPv4 the limited broadcast address is 255.255.255.255, and the
directed broadcast is a subnet's all-ones host address. At Layer 2, the broadcast MAC address is
FF:FF:FF:FF:FF:FF. Crucially, routers do not forward broadcasts, so they stay
inside the subnet or VLAN. Protocols like ARP ("who has this IP?") and DHCP
discovery rely on broadcast to find devices when addresses are not yet known.
Multicast: One-to-Many (a Group)
Multicast sends a single stream to a group of interested receivers. IPv4 multicast uses the
Class D range 224.0.0.0/4 (224.0.0.0 through 239.255.255.255). Hosts join and leave
groups using IGMP, and switches run IGMP snooping to forward the stream only
out ports with members instead of flooding it everywhere. Routing protocols use it too — OSPF, for
instance, sends hellos to 224.0.0.5 and 224.0.0.6. It is the natural fit for IPTV,
live streaming, and stock tickers.
Why Multicast Is So Efficient
The efficiency comes from replication by the network, not the source. To reach 100 receivers with unicast, the source sends 100 copies — saturating its own uplink. With multicast, the source sends one stream; routers and switches duplicate a packet only at the points where paths to different receivers diverge, so most links carry just a single copy. Unlike broadcast, multicast never bothers hosts that did not join, so it scales to large audiences without flooding the whole network.
Each Mode in the Real World
Every one of these modes is working on your network right now. Here is where you will actually meet them:
Unicast: Loading a Web Page
When you open a website, your browser opens a unicast connection to one server's IP address, and the server replies only to you. Email, SSH, and file downloads work the same way. If a million people load the page, that is a million independent unicast conversations — precise, but heavy on the server.
Broadcast: ARP and DHCP
Before your laptop can talk to the router, it broadcasts an ARP request — "who has
192.168.1.1?" — to every host on the subnet. A new device with no IP yet broadcasts a
DHCP discovery to find a server. Both rely on reaching everyone locally, and both stay
inside the subnet because routers never forward them.
Multicast: IPTV and Live Video
A TV provider streaming live channels to thousands of set-top boxes uses multicast: one stream per channel, replicated by the network only toward the boxes tuned in. Sending separate unicast copies to every viewer would melt the uplink. Stock tickers and financial market feeds use multicast for the same reason.
Common Misconceptions
These three modes trip up beginners in predictable ways. Keep these straight:
"Broadcasts Reach the Whole Internet"
They do not. A broadcast stays inside its broadcast domain — the local subnet or VLAN — because routers deliberately do not forward it. That is exactly why every home network can use broadcast-based protocols like ARP and DHCP without flooding the entire internet.
"Multicast Is Just Broadcast"
Multicast only reaches hosts that joined the group via IGMP; broadcast hits everyone whether they want the traffic or not. Multicast also crosses subnets (with multicast routing), while broadcast never leaves its segment. Treating them as the same thing leads to network designs that flood traffic needlessly.
"IPv6 Broadcasts Like IPv4"
IPv6 has no broadcast at all. It replaces broadcast with multicast — for example, the
all-nodes group ff02::1 reaches every node on the link, doing the job IPv4 broadcast used to do.
Expecting an IPv6 "broadcast address" will only cause confusion; reach for the right multicast group instead.
Addresses and How Hosts Join
Each mode has its own addressing. Unicast uses a device's normal IP and MAC. Broadcast
uses 255.255.255.255 (limited) or a subnet's all-ones host address (directed) at Layer 3, and
FF:FF:FF:FF:FF:FF at Layer 2. Multicast uses the Class D range 224.0.0.0/4;
the 224.0.0.0/24 block is reserved for link-local control traffic like routing protocol hellos, while
the 239.0.0.0/8 block is set aside for private, organization-local use. Hosts signal interest with
IGMP membership reports, and switches use IGMP snooping to avoid flooding the
stream to ports that have no members — otherwise a switch would treat multicast like a broadcast.
Why IPv6 Dropped Broadcast Entirely
IPv6 took the lesson that broadcast is wasteful — it interrupts every host's CPU even for traffic they do
not care about — and removed it completely. Everything broadcast used to do is now handled by well-scoped
multicast groups: ff02::1 for all nodes on the link, ff02::2 for all routers, and
solicited-node multicast groups that let Neighbor Discovery (the IPv6 replacement for ARP) reach
just the one host likely to own an address, instead of bothering the whole segment. The result is a quieter,
more efficient local network.
The bottom line: pick the mode that matches your audience. Talking to one device? Unicast. Need every local host to hear you, addresses unknown? Broadcast. Reaching a chosen group of subscribers efficiently, even a large one? Multicast — one stream, replicated only where it must be, and never bothering anyone who did not ask.
Frequently asked questions about unicast, broadcast, and multicast
What is the difference between unicast, broadcast, and multicast?
They are three ways to deliver packets. Unicast is one-to-one: a single sender to a single specific receiver, which is most internet traffic. Broadcast is one-to-all on a local network: delivered to every host in the broadcast domain. Multicast is one-to-many: a single stream is sent once and the network replicates it only toward the receivers who have joined a particular group.
What is a broadcast domain?
A broadcast domain is the set of devices that receive a broadcast frame sent by any one of them. It is bounded by routers, because routers do not forward broadcasts. A single VLAN or subnet is one broadcast domain, so a broadcast like an ARP request stays within that segment and never crosses to another subnet.
How do hosts join a multicast group?
On IPv4, hosts join and leave multicast groups using IGMP (Internet Group Management Protocol). A host sends an IGMP membership report to say it wants traffic for a group address such as 239.1.1.1. Switches use IGMP snooping to watch these messages and forward the multicast stream only out the ports with interested receivers, instead of flooding it to every port.
Why is multicast more efficient than unicast for streaming?
With unicast, the sender must transmit a separate copy of the stream to each receiver, so reaching 100 viewers means sending 100 copies. With multicast, the sender transmits a single stream and the network only replicates it where paths to different receivers diverge. That dramatically reduces the load on the sender and the links near it, which is why multicast is ideal for IPTV and live video.
Does IPv6 have broadcast?
No. IPv6 has no broadcast at all. It replaces broadcast with multicast, using well-known multicast groups for the jobs broadcast used to do. For example, the all-nodes multicast address ff02::1 reaches every node on the local link, taking the place of the IPv4 broadcast address. This is more efficient because uninterested hosts can ignore groups they have not joined.
Try the switch loop lab
You just saw how unicast, broadcast, and multicast reach different sets of hosts. The lab has no multicast, but it lets you trace unicast pings and ARP broadcasts across three looped switches, and see why real switches need spanning tree to stop one broadcast circulating forever.
Launch the switch loop lab →