IP has a sub range of its address space reserved for multicast addresses. In IPv4, these addresses are assigned in the class D address space, and IPv6 also has a portion of its address space (see Table 4.11) reserved for multicast group addresses. Some sub ranges of the multicast ranges are reserved for intra domain multicast, so they can be reused independently by different domains.
Thus, there are 28 bits of possible multicast addresses in IPv4 when we ignore the prefix sharedby all multicast addresses. This presents a problem when attempting to take advantage of hardware multicasting on a LAN. Let’s take the case of Ethernet. Ethernet multicast addresses have only 23 bits when we ignore their shared prefix. In other words, to take advantage of Ethernet multicasting, IP has to map 28-bit IP multicast addresses into 23-bit Ethernet multicast addresses. This is implemented by taking the low-order 23 bits of any IP multicast address to use as its Ethernet multicast address, and ignoring the high-order 5 bits. Thus, 32 (25) IP addresses map into each one of the Ethernet addresses.
When a host on an Ethernet joins an IP multicast group, it configures its Ethernet interface to receive any packets with the corresponding Ethernet multicast address. Unfortunately, this causes the receiving host to receive not only the multicast traffic it desired, but also traffic sent to any of the other 31 IP multicast groups that map to the same Ethernet address, if they are routed to that Ethernet. Therefore, IP at the receiving host must examine the IP header of any multicast packet to determine whether the packet really belongs to the desired group. In summary, the mismatch of multicast address sizes means that multicast traffic may place a burden on hosts that are not even interested in the group to which the traffic was sent. Fortunately, in some switched networks (suchas switched Ethernet) this problem can be mitigated by schemes wherein the switches recognize unwanted packets and discard them.