Windows Server 2008 : Network Addressing (part 1) - Addressing and Subnetting IPv4

1. Addressing Techniques

For the most part, you have used three different types of IP addressing techniques: automatic private IP addressing (APIPA), static, and dynamic (via DHCP).

1.1. APIPA

APIPA addressing is a Windows default mechanism for assigning IP addresses when a DHCP server is unreachable. This means that, no matter what the situation, machines running Windows will always have a logical address available to them within the 169.254.X.X range. For the most part, other than trivia, this addressing scheme is used at the MCITP level only as a reference to machines that are not communicating properly with DHCP.

1.2. Static

Static, or manual, addressing is the process of manually assigning an IP address to a machine based on a design created by an individual engineer or administrator. If network engineers could have their way, chances are that all IP addresses would be static. Unfortunately, in the modern day, that simply isn't possible because of the sheer number of addresses that have to be assigned.

1.3. Dynamic

Dynamic addressing is a technique that takes advantage of the Dynamic Host Control Protocol (DHCP) role that can be added to Windows Server. DHCP then automatically assigns addresses to requesting client machines through a predetermined pool within your DHCP server defined by the administrator. At the enterprise level, this is normally the most heavily used and implemented standard because of the ease, flexibility, and relatively equal efficiency of its addressing methods.

2. Address Ranges

When designing a network, the first step is to specifically establish the overarching breadth of what your design will need to encompass. This can vary wildly based on budget, number of users, addressing conventions (IPv4 or IPv6), and expectations of growth. At the top level, you first decide the addressing scheme you're going to use based on your needs within an address range.

2.1. IPv4 Address Ranges

Within a few years, the old method of IP address class ranges will become obsolete. With the advent of Internet Protocol version 6 (IPv6), the Internet now has simply so many different addressing assignments available on both the network and host levels that it's unlikely we'll run out any time in the near future, but stranger things have happened. Regardless, for the next few years, Internet Protocol version 4 (IPv4) will remain the most commonly used addressing assignment system in use.

IPv4 uses a set of four octets to create an individual, but not necessarily unique, logical address that can be used for the purposes of routing packets across networks. This is then further defined by a subnet mask, which partitions the address into different subnets for the purpose of sending and receiving broadcast traffic. At the top level, IP addresses are divided into five different classes that use a certain amounts of bits in the subnet mask for the network portion of your network and a certain amount of bits for your various hosts. It's rare that you'll discover a network administrator who uses all five classes of IPv4 addresses. For the most part, you are concerned with three different class levels of IP addresses: Class A, Class B, and Class C, which are described in Table 1. Each of these addressing classes has its own strengths and weaknesses, in that each can assign only a certain number of IP addresses based on the number of available host bits in the subnet mask.

Furthermore, each of these classes of networks is assigned certain ranges that will be predefined for your network design. Given your address class, you will fall into one of the ranges of Table 2.

When designing an addressing scheme, choosing (or discovering) your address class is your very first step. You need to determine both how many networks you require in your infrastructure and, perhaps more important, how many users are contained within these networks. Once you have determined this, you can then begin the process of subnetting your network.

2.2. IPv6 Address Ranges

Unlike its younger brother, IPv4, IPv6 no longer uses address classes. Instead, IPv6 uses prefixes that are subdivided by geographic locations around the world. Within those regions, the addresses are then subdivided more and more until its gets down to the individual level. In effect, this removes the need for the old fallback of the IPv4 addressing system, Network Address Translation (NAT). By design, IPv6 allows for every individual computer to theoretically have both a unique MAC address and a unique logical IP address, because so many addresses are available. Unlike IPv4, IPv6 uses eight quartets, making for a total of 128 bits worth of addressing space available.

3. Addressing and Subnetting IPv4

3.1. Working with the Number of Hosts and Subnetworks

In the past, Microsoft has asked lots of general subnetting questions of students just to make sure they understand the impact of choosing a particular subnet. The way that this is done is by examining both the host and network portions of the subnet mask and using a little mathematics to determine the number of available hosts and subnets that are possible.

For instance, you may be asked something akin to this: "If you're using a Class C address and you need to make sure your highly interactive network applications do not broadcast to more than five computers on the subnet, what would be the most appropriate mask to use?"

Or, building on the last example, you might be asked this: "Now that you have determined the amount of host bits required for your particular subnet, will you be able to accommodate 30 different subnetworks?"

In the following sections, you'll explore each of these questions and see whether you can find an easy way to calculate both of these answers. You'll start with determining the number of host bits.

3.1.1. Calculating Host Bits and Determining the Subnet Mask

In this scenario, you've been given a design requirement that no more than six computers exist on the same network (one application computer + five additional computers). Remember, the primary goal of a router is to stop broadcast traffic. And, in order to communicate from one subnet to another, you have to use a router. Thus, you have to determine the appropriate subnet mask to accommodate four hosts.

The following is the default subnet mask for a Class C address:

11111111.11111111.11111111.00000000

The part in bold is referred to as the host portion of the subnet mask. And the eight 0s refer to eight individual bits. The way you can determine the number of hosts is by taking 2, raising it to the number of available host bits, and then subtracting 2 again to give room for the broadcast address and the network address. In this case, the amount of available hosts would be 2⁸ − 2 = 254.

So, with that in mind, let's look at the example where you need four hosts per subnet. The easiest way to do this is by seeing which power of 2 gives you at least two more addresses than you need and then use that. You do this to accommodate for the broadcast and network addresses. Here are the powers of 2:

2¹ = 2

2² = 4

2³ = 8

2⁴ = 16

2⁵ = 32

2⁶ = 64

2⁷ = 128

2⁸ = 256

As you can see, in order to accommodate at least two more than 4 (in this case four more), you need to use 2⁁3 bits. In total, this gives you 2³ – 2 = 6 hosts.

So if you need three host bits to get six hosts, your network portion can take up all but 3 bits! That means the subnet mask for this problem is as follows:

11111111.11111111.11111111.11111000

The network portion is as follows:

11111111.11111111.11111111.11111000

And the host portion is as follows:

11111111.11111111.11111111.11111000

In shorthand, you would write this mask as /29, because it uses 29 bits. But the actual value for this mask is 255.255.255.X, where X is the last octet in decimal. And the number 11111 in decimal happens to be 248. Thus, the mask for this problem is as follows:

255.255.255.248

3.1.2. Determining the Number of Subnetworks

The next portion of the problem asks you to determine whether you can accommodate 30 different subnetworks. This can be easily answered by examining the number of subnet bits available to you. Since you're using a Class C address, you look at the network portion of the last octet only:

11111111.11111111.11111111.11111000

This is referred to as the subnet bits portion of the network portion of the subnet mask (that's a lot to say at once). But in reality, it's pretty easy to understand. This size grows and shrinks based on two factors. First, there is the address class. If you were using a Class A address, you'd look at all the networks bits in the second, third, and fourth octet combined. And if you were using a Class B address, you'd look at the third and fourth combined.

But since you're looking at a Class C address, you can just count those last five subnet bits. That is, you look at the 11111 portion of the 11111000 octet. And then, since you know it takes 5 bits, you can just raise 2 to the number of host bits (5) and derive the answer: 2⁵ = 32.

Thus, you have 32 subnetworks available, so, yes, you can accommodate the design requirements!

3.2. Addressing a Given Topology

The most common way you might be required to subnet on the MCITP exam is by being given a standard topology and then being asked to subnet the topology using only public addresses. The purpose of this exercise is to prove that you both understand how to subnet and can effectively design a complex infrastructure using Windows Server 2008. Normally, you will be given a diagram like Figure 1.

Figure 1. Publicly addressing a given topology

In the network diagram in Figure 2.1, your topology requires six different subnetworks because your overall design contains six different sites. Of course, from a Windows administrator point of view, this means you're going to have a lot of work on your hands: you have to subnet the network; connect the different subnets with site links; verify your replication strategy, schedule, and connections; and then do a little quality assurance to make sure everything is functioning properly. For now, you will stick to the initial process of subnetting the network. Let's assume for the moment that you have been given the Class C IP address range of 209.81.3.0.

Remember, this address is Class C because the first octet is greater than 192. This also means there is only a single address, but you require six different subnetworks because of the earlier requirements. To support this many networks, you have to do a minor calculation with a portion of your subnet mask, the last octet. Normally, the default subnet mask for a Class C network is as follows:

111111111.11111111.11111111.00000000

Or, written in decimal format, it's simply 255.255.255.0. Now, because you're operating within the boundaries of a Class C address, you're concentrating only on the last octet: 00000000.

The first objective, as you might imagine, is to make sure you have enough space in your subnet mask to support the amount of networks you need to accomplish your objective. To do this, you simply take the number of networks you need (six in this case) and go up the binary digits of binary notation until you reach a number greater than your original number:

So, you had to travel down four digits in binary in order to reach a number greater than six, your required amount of networks. Now, because you have this number, you can allocate the network portion of your subnet mask! Remember, the first bits of the subnet mask are the network portion, and the second bits are the host portion. So, instead of your default mask, shown in Figure 2, you move 4 bits in to subdivide your network, as shown in Figure 3.

Figure 2. Default mask

Figure 3. Subdivided mask

And thus, you now have a subnet mask for the entire network! Pretty cool, huh? However, this isn't the end of the process.

Whenever you subdivide a network into different portions, you have to understand what the range of your IP addresses is. This is because, naturally, a subnetted network cannot communicate with an IP address that is out of the range of its own subnet without a router. Finding these address ranges is actually a pretty simple process. You just take the last number of your subnet mask and see what it corresponds to in decimal notation.

In this case, you take the value shown here:

11111111.11111111.11111111.11110000

This value of 1 in the fourth spot in the mask is actually 32 in decimal notation. What this number means to you is that by starting at zero in the octet of concern (the first octet for Class A, the second octet for Class B, and the third octet for Class C), you add by 32 until you reach the number 256, which is outside the scope of a single octet of numbers. This will give you the network addresses (or network IDs) for the available subnetworks.

This is a lot easier to understand if you see it. All it means is that you can take your starting address and find your network identifiers by starting at 0 and adding 32 like this:

209.81.3.0

209.81.3.32 (0 +32)

209.81.3.64 (32 + 32)

209.81.3.96 (64 + 32)

209.81.3.128 (96 + 32)

209.81.3.160 (128 + 32)

209.81.3.192 (160 + 32)

209.81.3.224 (192 + 32)

Then, after you have your network identifier, you can find the broadcast address (the address that allows information to be sent to all devices within the subnet) by subtracting 1 from the last octet in all these numbers, with the exception of 0:

209.81.3.31 (32 – 1)

209.81.3.63 (64 – 1)

209.81.3.95 (96 – 1)

209.81.3.127 (128 – 1)

209.81.3.159 (160 – 1)

209.81.3.191 (192 – 1)

209.81.3.223 (224 – 1)

209.81.3.254 (256 – 1)

Then, you can find your usable addresses by looking at all the addresses between! You can see this illustrated in Table 3.

Most network administrators will steer you away from using the first and last subnet ranges, called the subnet zero and all-ones subnets. The reason behind this is that in the all-ones subnet, there can be confusion because you have a subnet with an identical broadcast address. Using the subnet-zero subnet was discouraged because you could have a subnet that was just 0. So, imagine seeing an IP address like 172.16.1.11 and having it be in the 172.16.0.0 subnet. Therefore, in practice, most administrators will say there are 2^N-2 usable subnets.

However, the strategy I have just discussed is an exceptionally powerful tool, and truth be told, you will be able to use this technique on more than just your MCITP certification exam. Many, many certifications use this technique, and if you've already mastered it, you're well on your way to attaining even more impressive titles.

Now, take a look back at Figure 2.1 again. Now that you have your network IDs, broadcast IDs, and usable addresses, you can assign each of these sites to a subnetwork and then start assigning them accordingly. If you'd like to see this in action, check out the "Readdressing a Network" sidebar.

Real World Scenario: Readdressing a Network You have just recently become employed with MegaCorp, a multibillion-dollar corporation that has placed you in charge of approximately 100 users within its organization. Currently, your portion of the organization is broken down into three separate networks connected by WAN connections. Today, the head office has decided it wants to readdress your network with the given address space of 209.113.60.0/27. Specifically, MegaCorp wants to make sure the fewest number of possible subnetworks is used but that each of these sustains enough host bits to support the required number of users. Given the network topology shown here, what subnet mask would you need to apply for the entire network, and what three separate broadcast addresses would you need to assign, assuming that the corporate specifications for network design require the lowest incremental broadcast address to be applied to Site A, then the next highest to Site B, and so forth? Growth is not a consideration in this design. Answer: Since you have no network with more than 30 users, you know you will need to reserve only enough host bits to maintain 30 users. Additionally, you have been told that growth is not an issue. The number 30 in binary is 00011110. Thus, when you look at this number, you realize that you require 5 bits for your host portion and 3 bits for your network portion. Thus, your default mask assigned by the head office of /27 will be appropriate. This is because the default mask of /24 for Class C, plus the additional 3 bits for your specific requirements, creates a need for a /27 subnet mask. However, to complete the design, you must define your broadcast addresses. These are discovered by taking the incremental number of the network, which is found by taking the binary value of the last bit in the network portion, incrementing the IP address by that value, and then subtracting zero. In your case, your mask is as follows: 11111111.11111111.11100000 And the binary value of the 1 at the 27th spot is 32. Thus, your incremental value is 32, giving you this as your network addresses: 209.113.60.0 209.113.60.32 209.113.60.64 209.113.60.96 This allows you to subtract 1 from 32, 64, and 96 for your broadcast address: Site A broadcast = 209.113.60.31 Site B broadcast = 209.113.60.63 Site C broadcast = 209.113.60.95

3.3. VLSM

Variable-length subnet mask (VLSM) is a technology concept that is used throughout networks that have great subnetting requirements but require the least amount of waste and most efficient use of broadcast traffic possible. Imagine for a moment that you work for MyCorp again and that you have a user situation in your office that breaks down as shown in Figure 4.

As you can see, the network is broken up into four individual subnets, and each of those subnets has a different user requirement. In the first subnet, you require 200 users, in the second you require 30, and so forth. Normally, you would break this network up by subnetting it into four individual groups, Subnet A, B, C, and D. But let's look at what happens if you do that; see Figure 5.

On the right side of the figure, you'll see the maximum number of hosts that can be contained on the subnet. To the right of this, you'll see your actual number of users, and then on the far right you can see the amount of wasted addresses within this space. Now, as you can imagine, especially with high-end networking, space matters.

Figure 4. Networks with various user requirements

Figure 5. Standard subnetting applied

Part of being an enterprise administrator is understanding that your ultimate goal for your enterprise is to make sure that everything is running as often as possible and as efficiently as possible. In your career up until this point, someone has undoubtedly told you that the best way to become successful in life is not to waste anything. This includes money, opportunities, and, of course, host addresses. Instead, consider what would happen if you could apply an individual subnet address for each of these subnetworks. For instance, using some basic calculations that you learned earlier, you could determine that the following masks could be used based on the host requirements:

This sure would be nice if you could do it, wouldn't it? The truth is, you can. VLSM allows administrators to apply a more specific subnet mask to a preexisting subnet mask to further define a subnet. This is extraordinarily useful for complex organizations and can really optimize your efficiency. Consider your first example, where the same mask was applied throughout, and consider your latest example, shown in Figure 6, which uses VLSM.

Figure 6. VLSM vs. standard subnetting

The numbers speak for themselves.

Now, if you wanted to apply these addresses to your network, you'd have to go one step further and calculate the address ranges where you would need to apply them. First you'd start by taking 2 and raising it to the required number of bits per individual subnet mask to find out an incremental number like you did earlier. Now, you've managed to subnet your network into several different areas and saved yourself tons of IP addresses!