Windows Server 2008 : Network Addressing (part 2) - Addressing IPv6

4. Addressing IPv6

So, if you thought subnetting IPv4 was confusing, wait until you meet its younger brother, IPv6! That's actually only a joke. In truth, addressing IPv6 isn't as bad as you would think. But when most people even look at IPv6 addresses, they immediately assume they must be inherently evil because of their obvious unreadability. If you look at an IPv6 address like 4305:A93E:BADC:8956:3586:8D9C:7032:1423, it looks like garbage. You may know instinctively upon seeing it that it's been laid out in IPv6 format, but it just seems like a lot of random numbers and letters thrown together in one place. Thankfully, there's a lot of reason behind the seeming randomness.

4.1. IPv6 Shorthand Notation

One of the quickest ways you can identify and evaluate an IPv6 address is by using its shorthand notation. As you can tell, a full-length IPv6 address is very long. But thankfully, most of the time, you will see a lot of zeros in an IPv6 address. A very realistic example of one you may encounter is the reserved multicast address FF02:0000:0000:0000:0000:0000:0001:0002.

As you can see, this particular address has a seemingly excessive amount of zeros (25, in fact). Accordingly, you can shorten this by using the :: notation, which essentially means "use zeros until." When you see this symbol start, from that point on, you can insert zeros until you reach a number. Then, after this number, you can see how many sets of single : have been used to see what octet it represents. And just to make things even easier, if an octet is preceded by a few zeros, such as the octet 0002, you simply write this as 2 behind the : symbol. This sounds complicated, but it's actually pretty easy. The previous example in shorthand would be FF02::1:2.

In plain English, this means "FF02 is my first octet; then I keep going until I reach an octet that ends in 1, and then my last octet has a 2." Let's try it—it will probably make sense if you do it one part at a time by using a simple step-by-step procedure:

4.2. Anatomy of IPv6

IPv6 addresses are beautiful because of their absolute simplicity. Before, when you dealt with an IPv4 address, there was a lot of confusion. What part of the address belongs to the Internet service provider? Where is the subnet portion of the address? Better yet, where is the host? In IPv6, these are no longer concerns.

All IPv6 addresses can be broken down into two distinct portions, which can further be subdivided to a point that just about every portion of the address is accounted for. On the base level, IPv6 addresses are broken into two 64-bit portions, one of which is called the prefix portion and one of which is called the host portion, or the interface ID. Visually, the address looks like Figure 7.

Figure 7. IPv6 address portions

In one fell swoop, you can cover the second portion of the address. It's just the host portion of the network. In more technical terms, the 65th to the 128th bit of your address is completely dedicated to assigning the address to your hosts. That's a lot of hosts! It's more, in fact, than even some of the largest enterprises on the planet would ever use. However, when the IEEE designed IPv6, it didn't want to run into a situation where anyone would ever have to worry about having "enough" host addresses ever again. I think it's safe to say they've succeeded. 2⁁64 is such a large number that if you were to take that many pennies and stack them up one after another, you'd be able to reach Mars more than 300,000 times. Or, if you'd like to think of it in more Microsoft terms, you'd be able to have 230,584,300 times the amount of money of Bill Gates (when he was worth 80 billion).

The first portion of an IPv6 address, called the address prefix, is a little bit more complicated, but not too much so. To begin, one of the real issues that IPv6 was meant to fix was to give service providers their own reserved section of the IP address that would identify whatever service provider was issuing the address. Accordingly, the IEEE assigned the first 48 bits of the prefix portion of the address to the service provider. Then, with the remaining 16 bits, it allocated a portion to be used for subnet addressing. You can see another visual interpretation of this in Figure 8.

Figure 8. IPv6 complete address portions

The main reason that only 16 bits has been assigned to the subnet portion is actually pretty reasonable. After all, how often do you run across an organization that will need more than 65,536 subnets? The answer is not very often. And thus, only a small portion of the overall 128 bits is assigned. In just a moment, I'll go over how subnetting this portion of an address is slightly different than it was with IPv4. But for the moment, I'll take a step back and talk about those first 48 bits before the 16 bits of the subnet portion.

There are three organizations that take a bite out of the first 48 bits of addresses. These are the ICANN, RIR, and ISPs:

• Internet Corporation for Assigned Names and Numbers (ICANN)

• Regional Internet Registry (RIR)

• Your Internet service provider (ISP)

Thankfully, the exact scope of the importance of these organizations is outside the objectives of this exam. Suffice to say, the Internet address prefix goes through three filters going from ICANN to RIR to ISP that more and more uniquely define the coverage area of these addresses.

4.3. IPv6 Address Types

One of the biggest changes that came with IPv6 was the complete and total removal of the concept of a broadcast address. And if you ask most busy administrators, that's a good thing. The reason is that IPv6 has instead replaced the need for broadcast addresses with the concept of multicast addressing. The word multicast is getting a little ahead of myself, so I'll start by defining the three different types of addresses that are available to you in IPv6:

Unicast

A unicast address is an address that is assigned to a particular host so that host, and only that one particular host, can send and receive data. It's equivalent to saying "You and only you are identified as this."

Multicast

A multicast address is effectively a grouping of addresses that is addressed for the point of sending and receiving information to that group. So, if you wanted to send a broadcast of information, you could send it to a particular multicast group.

Anycast

The name is a bit confusing, but an anycast address is similar to a multicast address in that it's sent to a particular group of addresses, but only the address "nearest" to it. So, instead of sending it to every member of the group, it sends to a particularly near member of that group.

For the purposes of this book, I'll concentrate on unicast addresses, because they're what you as an administrator most care about. Concepts such as multicast addressing are more designed toward network administration and engineering, because they determine what particular routing protocols are used and where they can and cannot be broadcast throughout the network.

4.4. IPv6 Static and Dynamic Addressing

Just like IPv4, IPv6 addresses can be assigned both dynamically and statically. If an administrator for some reason wants to assign a particular address to some device within their network, they most certainly may. Furthermore, there are plenty of new methods to dynamically assign IPv6 addresses to devices through the use of Dynamic Host Control Protocol version 6 (DHCPv6).

However, there are a few differences to which you need to pay careful attention. In particular, you need to be interested in the conventions IPv6 uses to effectively assign addresses throughout the entire network. In total, there are four possible combinations—two of which are used for static addressing and two of which are used for dynamic. I'll talk about the static methods and dynamic methods.

4.4.1. EUI-64

One of the great benefits of having such an incredibly long host field is not only the ability to have an absolutely gargantuan number of hosts but also the ability to specify a great deal of uniqueness toward an individual address. As you're familiar with from your study of basic networking, an individual interface normally contains two addresses, a logical Internet protocol address (IP) and a physical MAC address. In IPv4, the MAC address happened to be larger than the IP address. That is, IPv4 addresses were only 32 bits in length, but MAC addresses were 48 bits in length. The original purpose for this design (and still the purpose to this day) is twofold. First, a MAC address specifies a unique physical address for your computer. Second, it provides an address that a switch can use to forward a frame. Just in case you haven't seen one in a while, a typical MAC address looks like this:

00-1A-A0-05-2A-B7

Normally, a MAC address is divided into six different sets of two hex numbers for readability; let's do something different for a moment and split the example address into two sets of six hex numbers. The reason why will become clear in a moment.

001AA0 052AB7

Now that you've separated these two values, let's shift gears for a moment. Remember earlier when you read that IPv4 addresses were smaller than MAC addresses? Well, that simply isn't the case with IPv6. In fact, just the host portion alone is 16 bits larger than the entire MAC address.

Accordingly, a few networking geniuses decided it would be really fun (and really easy) to sort of semi-use the MAC address in the host field. It gives a unique address, and to boot, it allows static addressing without the need to manual enter every single number.

I say that the address is "semi-used" because in order to complete the 64-bit host fields, you're lacking 16 bits. Thus, you need to insert 16 bits somewhere in the host field to make up for this lack of bits. To do this, you use the hex field:

FFFE

Then, just to establish a little uniqueness (and for a few more technical reasons that are beyond the scope of this book), the seventh bit of the MAC address is flipped. So, for this example, the address is as follows:

001AA0052AB7

To achieve 7 bits, you need only the first two values (00). Thus, you take those first two hex numbers and convert them into binary:

00000000

And then, you "flip" the seventh bit:

00000010

In hex, this value comes out to 02. Thus, your new address is as follows:

021A:A0FF:FE05:2AB7

Visually, you can think of it like Figure 9.

Figure 9. EUI-64 visual breakdown

4.4.2. Manual Assignment

The second way an address can be assigned statically in IPv6 is by doing it the old-fashioned way. And, although it may be a lot more tedious to implement, it's certainly more easily explained. Just like in IPv4, you can manually punch in an address piece by piece. The only real difference is that one takes decimal notation and the other takes hexadecimal notation and a subnet prefix. You can see the Windows Server 2008 manual assignment dialog box in Figure 10.

It may be old-fashioned, but it still works!

Figure 10. Windows Server 2008 manual IPv6

4.4.3. DHCPv6

In DHCPv6 there are two supported states of DHCP: stateful and stateless. Stateful DHCP is similar to what you've experienced in the past with DHCPv4; it just means that DHCP tracks the state of the interfaces it communicates with, such as information regarding the client and how long the lease on the dynamic address exists. The only real difference is that DHCPv4 uses broadcasts in order to find a DHCP. Clients, when first connected, essentially advertise themselves on their subnet by saying "Here I am!" And then the DHCP server responds accordingly. Although this works fine for DHCPv4, unfortunately DHCPv6 doesn't use broadcasts. So, it sets aside a default multicast address that I told you you'd probably see sometime in the future. That address is the following:

FF02:0000:0000:0000:0000:0000:0001:0002

In stateless DHCP, the "state information" (whether an interface is up or down, how long the lease exists, and so on) is ignored. Typically, stateless DHCP is used in conjunction with stateless autoconfiguration, which is a method used by IPv6 to automatically assign addresses to given interfaces based on their EUI-64 address. The main difference between stateless and stateful is that stateless doesn't remember IP addresses, but it can still supply information such as a DNS server.