Unicast is an addressing and delivery method that sends traffic from one sender to exactly one receiver by placing a destination address in the IP packet header, allowing routers to forward the packet along a single path to that specific interface. In enterprise networks, unicast carries most traffic including web requests, DNS queries, and DHCP renewals, making accurate IP address management critical since every unicast destination must be a unique address within its scope to ensure traffic reaches the correct host. Poor management of unicast addressing leads to operational problems such as duplicate addresses, overlapping subnets, and unreachable hosts, while proper unicast implementation enables security teams to apply per-host access controls and logging based on identifiable source and destination addresses.
How unicast works
In unicast delivery, the sender places the address of one destination interface in the packet header. Each router along the path makes a forwarding decision based on that destination address.
- Address the packet: The sending host sets its own address as the source and the address of the intended receiver as the destination in the IP header.
- Choose the next hop: The host sends the packet directly to the receiver if it is on the same subnet. Otherwise it sends the packet to its default gateway.
- Forward hop by hop: Each router looks up the destination in its routing table, using the longest matching prefix, and passes the packet to the next hop.
- Deliver to the interface: The last router resolves the destination to a link-layer address, and the receiving interface accepts the packet because the address matches its own.
IPv6 defines several unicast address types in RFC 4291, including global unicast addresses, which are routable on the internet, and link-local addresses, which are valid only on a single link and use the fe80::/10 prefix. Unicast differs from multicast and broadcast, where one packet can reach many receivers.
Common misconceptions about unicast
Unicast describes the number of receivers, not the path length. A unicast packet may cross many routers, and each one makes its own forwarding decision.
Engineers sometimes expect IPv6 link-local addresses to work across subnets. They are valid only on a single link and routers do not forward them, so services must use a global or unique local address to be reachable elsewhere.
Addresses from RFC 1918 ranges are still unicast addresses. They are scoped to private networks, not a separate delivery type, and overlapping use across sites causes ambiguity until translation or renumbering is applied.
Uniqueness applies only within the address scope. The same IPv4 private address can exist on separate networks, and the same IPv6 link-local address can appear on different links, so tracking must record the scope as well as the address.
Sending the same data to many receivers over unicast duplicates traffic once per receiver. For large audiences such as video distribution or software updates, this consumes sender and link bandwidth that multicast or caching could reduce.