Chapter 17: IP Version 6

Cisco Press

1 2 3 4 5 6 Page 2
Page 2 of 6

As shown in the figure, the routers installed by ISPs in other major geographies of the world can have a single route that matches all IPv6 addresses in North America. While hundreds of ISPs might be operating in north America, and hundreds of thousands of enterprise customers of those ISPs, and tens of millions of individual customers of those ISPs, all the public IPv6 addresses can be from one (or a few) very large address blocks—requiring only one (or a few) routes on the Internet routers in other parts of the world. Similarly, routers inside other ISPs in North America (for example, NA-ISP2, indicating North American ISP number 2 in the figure) can have one route that matches all address ranges assigned to NA-ISP2. And the routers inside NA-ISP1 just need to have one route that matches the entire address range assigned to Company1, rather than needing to know about all the subnets inside Company1.

Besides keeping the routers' routing table much smaller, this process also results in fewer changes to Internet routing tables. For example, if NA-ISP1 signed a service contract with another enterprise customer, NA-ISP1 could assign another prefix inside the range of addresses already assigned to NA-ISP1 by ARIN. The routers outside NA-ISP1's network—the majority of the Internet—do not need to know any new routes, because their existing routes already match the address range assigned to the new customer. The NA-ISP2 routers (another ISP) already have a route that matches the entire address range assigned to NA-ISP1, so they do not need any more routes. Likewise, the routers in ISPs in Europe and South America already have a route that works as well.

While the general concept might not be too difficult, a specific example can help. Before seeing a specific example, however, it helps to know a bit about how IPv6 addresses and prefixes are written.

Conventions for Representing IPv6 Addresses

IPv6 conventions use 32 hexadecimal numbers, organized into 8 quartets of 4 hex digits separated by a colon, to represent a 128-bit IPv6 address. For example:


Each hex digit represents 4 bits, so if you want to examine the address in binary, the conversion is relatively easy if you memorize the values shown in Table 17-2.

Table 17-2 Hexadecimal/Binary Conversion Chart





































Writing or typing 32 hexadecimal digits, while more convenient than doing the same with 128 binary digits, can still be a pain. To make things a little easier, two conventions allow you to shorten what must be typed for an IPv6 address:

  • Omit the leading 0s in any given quartet.

  • Represent 1 or more consecutive quartets of all hex 0s with a double colon (::), but only for one such occurrence in a given address.

NOTE - For IPv6, a quartet is one set of 4 hex digits in an IPv6 address. Eight quartets are in each IPv6 address.

For example, consider the following address. The bold digits represent digits in which the address could be abbreviated.


This address has two different locations in which one or more quartets have 4 hex 0s, so two main options exist for abbreviating this address, using the :: abbreviation in one or the other location. The following two options show the two briefest valid abbreviations:

  • FE00::1:0:0:0:56

  • FE00:0:0:1::56

In particular, note that the :: abbreviation, meaning "one or more quartets of all 0s," cannot be used twice, because that would be ambiguous. So, the abbreviation FE00::1::56 would not be valid.

Conventions for Writing IPv6 Prefixes

IPv6 prefixes represent a range or block of consecutive IPv6 addresses. The number that represents the range of addresses, called a prefix, is usually seen in IP routing tables, just like you see IP subnet numbers in IPv4 routing tables.

Before examining IPv6 prefixes in more detail, it is helpful to review a few terms used with IPv4. IPv4 addresses can be analyzed and understood using either classful addressing rules or classless addressing rules. (This book and CCENT/CCNA ICND1 Official Exam Certification Guide both use classful terminology for the most part.) Classful addressing means that the analysis of an IP address or subnet includes the idea of a classful network number, with a separate network part of the address. The top part of Figure 17-2 reviews these concepts.

Figure 17-2

IPv4 Classless and Classful Addressing, and IPv6 Addressing

Thinking about IPv4 addressing as classful addresses helps to fully understand some issues in networking. With classful addressing, for example, the written value means 16 network bits (because the address is in a Class B network) and 8 host bits (because the mask has 8 binary 0s), leaving 8 subnet bits. The same value, interpreted with classless rules, means prefix, prefix length 24. Same subnet/prefix, same meaning, same router operation, same configuration—it's just two different ways to think about the meaning of the numbers.

IPv6 uses a classless view of addressing, with no concept of classful addressing. Like IPv4, IPv6 prefixes list some value, a slash, and then a numeric prefix length. Like IPv4 prefixes, the last part of the number, beyond the length of the prefix, is represented by binary 0s. And finally, IPv6 prefix numbers can be abbreviated with the same rules as IPv4 addresses. For example, consider the following IPv6 address that is assigned to a host on a LAN:


This value represents the full 128-bit IP address; in fact, you have no opportunities to abbreviate this address. However, the /64 means that the prefix (subnet) in which this address resides is the subnet that includes all addresses that begin with the same first 64 bits as the address. Conceptually, it is the same logic as an IPv4 address. For example, address is in the prefix (subnet) whose first 24 bits are the same values as address

Like with IPv4, when writing or typing a prefix, the bits past the end of the prefix length are all binary 0s. In the IPv6 address shown previously, the prefix in which the address resides would be as follows:


When abbreviated, this would be:


Next, one last fact about the rules for writing prefixes before seeing some examples and moving on. If the prefix length is not a multiple of 16, the boundary between the prefix and the host part of the address is inside a quartet. In such cases, the prefix value should list all the values in the last octet in the prefix part of the value. For example, if the address just shown with a /64 prefix length instead had a /56 prefix length, the prefix would include all the first 3 quartets (a total of 48 bits), plus the first 8 bits of the fourth octet. The last 8 bits (last 2 hex digits) of the fourth octet should now be binary 0s. So, by convention, the rest of the fourth octet should be written, after being set to binary 0s, as follows:


The following list summarizes some key points about how to write IPv6 prefixes:

  • The prefix has the same value as the IP addresses in the group for the first number of bits, as defined by the prefix length.

  • Any bits after the prefix-length number of bits are binary 0s.

  • The prefix can be abbreviated with the same rules as IPv6 addresses.

  • If the prefix length is not on a quartet boundary, write down the value for the entire quartet.

Examples can certainly help a lot in this case. Table 17-3 shows several sample prefixes, their format, and a brief explanation.

Table 17-3 Example IPv6 Prefixes and Their Meanings



Incorrect Alternative


All addresses whose first 3 bits are equal to the first 3 bits of hex number 2000 (bits are 001)

2000/3 (omits ::)

2::/3 (omits the rest of the first quartet)


All addresses whose first 26 bits match the listed hex number

2340:114::/26 (omits the last digit in the second quartet)


All addresses whose first 32 bits match the listed hex number

2340:1111/32 (omits ::)

Almost as important to this convention is to note which options are not allowed. For example, 2::/3 is not allowed instead of 2000::/3, because it omits the rest of the octet, and a device could not tell whether 2::/3 means "hex 0002" or "hex 2000." Only leading 0s in a quartet, and not trailing 0s, can be omitted when abbreviating an IPv6 address or prefix.

Now that you understand a few of the conventions about how to represent IPv6 addresses and prefixes, a specific example can show how ICANN's IPv6 global unicast IP address assignment strategy can allow the easy and efficient routing shown back in Figure 17-1.

Global Unicast Prefix Assignment Example

IPv6 standards reserve the 2000::/3 prefix—which, when interpreted more fully, means all addresses that begin with binary 001 or either a hex 2 or 3—as global unicast addresses. Global unicast addresses are addresses that have been assigned as public and globally unique IPv6 addresses, allowing hosts using those addresses to communicate through the Internet without the need for NAT. In other words, these addresses fit the purest design for how to implement IPv6 for the global Internet.

Figure 17-3 shows an example set of prefixes that could result in a company (Company1) being assigned a prefix of 2340:1111:AAAA::/48.

Figure 17-3

Example IPv6 Prefix Assignment in the Internet

The process starts with ICANN, which owns the entire IPv6 address space, and assigns the rights to registry prefix 2340::/12 to one of the RIRs, ARIN in this case (North America). This means that ARIN has the rights to assign any IPv6 addresses that begin with the first 12 bits of hex 2340 (binary value 0010 0011 0100). For perspective, that's a large group of addresses—2116 to be exact.

Next, NA-ISP1 asks ARIN for a prefix assignment. After ARIN ensures that NA-ISP1 meets some requirements, ARIN might assign ISP prefix 2340:1111::/32 to NA-ISP1. This too is a large group—296 addresses to be exact. For perspective, this one address block might well be enough public IPv6 addresses for even the largest ISP, without that ISP ever needing another IPv6 prefix.

Finally, Company1 asks its ISP, NA-ISP1, for the assignment of an IPv6 prefix. NA-ISP1 assigns Company1 the site prefix 2340:1111:AAAA::/48, which is again a large range of addresses—280 in this case. In the next paragraph, the text shows what Company1 could do with that prefix, but first, examine Figure 17-4, which presents the same concepts as shown in Figure 17-1, but now with the prefixes shown.

Figure 17-4

IPv6 Global Routing Concepts

The figure shows the perspectives of routers outside North America, routers from another ISP in North America, and other routers in the same ISP. Routers outside North America can use a route for prefix 2340::/12, knowing that ICANN assigned this prefix to be used only by ARIN. This one route could match all IPv6 addresses assigned in North America. Routers in NA-ISP2, an example alternative ISP in North America, need one route for 2340:1111::/32, the prefix assigned to NA-ISP1. This one route could match all packets destined for all customers of NA-ISP1. Inside NA-ISP1, its routers need to know to which NA-ISP1 router to forward packets to for that particular customer (the router named ISP-1 in this case), so the routes inside NA-ISP1's routers lists a prefix of 2340:1111:AAAA/48.

Subnetting Global Unicast IPv6 Addresses Inside an Enterprise

The original IPv4 Internet design called for each organization to be assigned a classful network number, with the enterprise subdividing the network into smaller address ranges by subnetting the classful network. This same concept of subnetting carries over from IPv4 to IPv6, with the enterprise subnetting the prefix assigned by its ISP into smaller prefixes. When thinking about the IPv6 subnetting concept, you could make the following general analogies with classful IPv4 subnetting to help understand the process:

  • The prefix assigned to the enterprise by the ISP, which must be the same for all IPv6 addresses in one enterprise, is like the IPv4 network part of an address.

  • The enterprise engineer extends the length of the prefix, borrowing host bits, to create a subnet part of the address.

  • The last/third major part is the host part of the address, called the interface ID in IPv6, and is meant to uniquely identify a host inside a subnet.

For example, Figure 17-5 shows a more detailed view of the Company1 enterprise network shown in several of the earlier figures in this chapter. The design concepts behind how many subnets are needed with IPv6 are identical to those for IPv4: A subnet is needed for each VLAN and for each serial link, with the same options for subnets with Frame Relay. In this case, two LANs and two serial links exist, so Company1 needs four subnets.

The figure also shows how the enterprise engineer extended the length of the prefix as assigned by the ISP (/48) to /64, thereby creating a 16-bit subnet part of the address structure. The /48 prefix is generally called the site prefix, and the longer prefix used on each link is called a subnet prefix. To create this extra 16-bit subnet field, the engineer uses the same concept as with IPv4 when choosing a subnet mask by borrowing bits from the host field of an IPv4 address. In this case, think of the host field as having 80 bits (because the prefix assigned by the ISP is 48 bits long, leaving 80 bits), and the design in Figure 17-5 borrows 16 bits for the subnet field, leaving a measly 64 bits for the host field.

Figure 17-5

Company1 Needs Four Subnets

A bit of math about the design choices can help provide some perspective on the scale of IPv6. The 16-bit subnet field allows 216, or 65,536, subnets—overkill for all but the very largest organizations or companies. (There are also no worries about a zero or broadcast subnet in IPv6!) The host field is seemingly even more overkill: 264 hosts per subnet, which is more than 1,000,000,000,000,000,000 addresses per subnet. However, a good reason exists for this large host or interface ID part of the address, because it allows one of the automatic IPv6 address assignment features to work well, as is covered in the section "IPv6 Host Address Assignment," later in this chapter.

1 2 3 4 5 6 Page 2
Page 2 of 6
The 10 most powerful companies in enterprise networking 2022