The second network card acts as the default gateway for remote packets for the first network card and vice versa. The routing software then takes care of the actual packet transfer. In TPC/IP networks, the Network Mask is used to determine whether a packet should be routed. The packet is routed if the network number for the destination host does not match the network number of the local host.

The two networks are 222.123.1.0 and 222.123.2.0 and the server has two correctly installed network interface cards, NIC's, (proper IRQ, RAM, and PORT as required). You will use a default Class C Network Mask. The first NIC will be 222.123.1.10 and will use the second NIC as its gateway, 222.123.2.20. NIC 2 will be similarly set up with the IP numbers reversed.

Please note: You must be logged on with administrator privileges to correctly install all services in any of these exercises.

1. Go to the Network Icon in the Control Panel and Click to start the Network Control Screens.
2. Click on the Protocols Panel. This will display all installed protocols. You should see TCP/IP Protocol listed in this display. If not, you need to add the protocol first.
3. Click on the TCP/IP Protocol and then click on Properties.
4. In the Network Card drop-down List there should be two Network Cards numbered [1] and [2]. For the first network card, enter the following

IP Address: 222.123.1.10

Network Mask: 255.255.255.0

Gateway: 222.123.2.20

5. Click the drop-down box and click on the second network card. Enter the following:

IP Address: 222.123.2.20

Network Mask: 255.255.255.0

Gateway: 222.123.1.10

You should now see the following configuration changes:

222.123.1.0 à 222.123.1.10

Server: 10NT20

222.123.2.20 ß 222.123.2.0

6. Click on the Routing tab at the top.
7. In the Routing Tab panel click the Enable IP Forwarding box.
8. Click the OK button at the bottom to accept all changes made to the TCP/IP properties. This will return you to the Network Control window.

You could click Close if you are finished now or move on to the exercise to install RIP for Internet Protocol and then click on the Close button.

Single Adapter Systems

There is a special case of a single card acting as a router to two sets of networks that actually share the same local area cabling media.

In the case of the single network interface card you would have to have two distinct network IP addresses assigned to it, one for each network to be accessed. Static routes should be added to the interface so that packets from either network interface will use the other IP address on the same card as their default gateway. This is also known as a logically multihomed network interface.

Number of IP Addresses Assigned to an Interface

There is a limit of five IP addresses for any one network adapter, to make the sys admin’s job manageable. Have you ever had to take two older machines with separate functions, like FTP and HTTP protocol servers, and integrate them onto one machine? The new machine can replace the two older machines and have both services appear as before, with independent IP addresses even though they are the same network card.

Number of IP Gateways Assigned to an Interface

Additionally, you can assign alternate gateways to use if there are multiple routers providing routing to other networks. This provides failover should the primary gateway fail to forward the packet.

Advanced IP Options: Multiple IP and Gateway Addresses

To add additional IP and/or gateway addresses to one network interface, click the Advanced option, lower right corner, on the IP Address panel of the TCP/IP Properties control window. Figure 5-4 shows the Advanced Options input control sheet.

Only if your network has multiple gateways to access the remote networks and only if you need to have multiple IP addresses responded to by this machine, would you ever need to add these extra addresses. Simple installations will not need this, such as in the example network in Figure 5-1.

Static Versus Dynamic Routing

Early routers had to be programmed with exactly which networks they could route between which interfaces, especially if there were many network interfaces. This configuration method is called static routing. Pre-set routing tables are loaded from a disk file or EPROM into the router software at boot time or after any reset or reboot.

Dynamic routing requires additional software that sends special route information packets between routers. One of the early versions that is still used extensively is called the Routing Internet Protocol (RIP). NT provides this as a service you can add. (See the section on RIP in this chapter for more details.)

There are some newer routing information services called Border Gateway Protocol (BGP) and Open Shortest Path First (OSPF). They require router software that understands the packets and can use the information to determine the best, shortest, least cost, or alternate route to a remote network based on information passed back to the local router. You would need to add the Steelhead Routing upgrade to NT 4.0 to get this additional service.

Features of Static Routing

Static routing refers to a specific table of routes input by you. This requires you as the router manager to manually configure the specific routing information for each interface. NT 4.0 is considered a static router until the addition of software to make it a dynamic router.

Each network card can be programmed with explicit routes to allow and or disallow specific routes. This gives the administrator explicit control over each interface. Additionally, you as the administrator can add the costs associated with routes as well as other related metrics to manage the flow of traffic between networks.

Advantages

You can be very specific about the routing, filtering, and settings on every routing interface.

It is very difficult to corrupt from external sources and can be more secure than dynamic routing.

Disadvantages

You must manually set up every interface and each interface will be different.

On large networks with many routers, gateways, and alternate routes, you also need to add information about metrics such as cost, which is a metric value you establish to compare route usage, to manage the flow of packets.

Features of Dynamic Routing

Dynamic routing protocols advertise the routes they are familiar with and pass on the metrics, number of other routers, or hops, required to get from their host to another network, either directly or indirectly through another router.

Advantages

No further set up is required for all the interfaces, they learn from the other routers over time.

Disadvantages

Dynamic routing generates continuous traffic from the routers with route update information.

They can be attacked from external sources with falsified RIP packets, which requires much more security and attention by a security officer, or you, for corruption and attacks.

They can create circle routes in complex networks with more than 16 hops.

Comparing Static and Dynamic Routing

With just two networks, the static routing setup is the most appropriate. In Figure 5-1, the routing interfaces have one static route and their default gateways are the other interface card. This is a basic static route. There are no other routes or options.

If your network has three or four parallel networks, as illustrated in Figure 5-5, then static routing may still be useful, although dynamic routing would be easier to set up.

However the actual setup is far more complex. You cannot simply have the routers route to the other network by assigning the gateway addresses to the other network interface card. This simply moves the packet to the other side when in actual fact the router needs to send the packet on to the specific gateway from that network to the next network.

In Figure 5-5, the network 223.4.5.0, Net5, needs to route 223.4.6.0, Net6, packets to 223.4.6.253. But packets for Net7 and Net8 from Net5 need to be sent to the next gateway, 223.4.6.252, so they can be forwarded on their way. Same goes for every other router interface. You must program each interface as to where to send packets from every other network.

In Figure 5-5 the left-most network is numbered 223.4.5.0 and has hosts (223.4.5) .51, .52, and .53 plus the router interface .254 for a total of four interfaces. The right-most network also has four interfaces or nodes. The middle two networks have three hosts each and two router interfaces each.

Although the router is drawn as a single box, it may represent two servers with a dial-in connection using a modem from either side. The exact hardware is not important here, just the concepts first.

For the outside networks, 223.4.5.0 and 223.4.8.0, all clients are configured with their gateway pointing to the only router on their networks, 223.4.5.254 and 223.4.8.249, respectively. For these two outside networks, the router is their only gateway to all the other networks. A packet destined for any network other than the next one over requires that you program the router interface for a network one or two more hops away.

The middle two networks have a more interesting setup. They both have two router ports or gateways. You can program your client host machines with only one IP gateway address, so which do you pick? Does it matter? Either and maybe are the answers. You can pick either gateway and then program each gateway with static routes for the other remote gateways or pick the one most likely to be used most often. Secondly, it does not matter once proper additional routes have been added, although if the most heavily used gateway is selected, this will reduce the number of routers that will handle the packet.

Routing Tables

Each independent router port has a routing table. In Figure 5-1 with the two networks only, the routing table would be a simple one with one entry for each card pointing to the other network interface as the gateway. This is a simplified description, as there are other routes which are not discussed in detail, such as references to internal interfaces, network, and broadcast addresses. Some of these are seen in Table 5-1.

In Figure 5-4, all three routers and their associated six ports would need to be programmed with the required static routes of the other networks if dynamic routing was not being used. You would want to make these permanent entries as well so that when the NT system is rebooted, the routes are re-established.

Entries

At the command prompt is a utility called ROUTE that is used to view and manage routes that are static and dynamic for an interface. You can delete and change dynamic routes but the next dynamic update may refresh the information.

At a command prompt if you have installed TCP/IP services, you can enter:

ROUTE To get help on all the commands

ROUTE PRINT To display the current routes configured

Default

The default routes set up initially define some standard network numbers used internally and locally for test and broadcast.

Even on a basic network card interface there are numerous routes. Table 5-1 shows the output from a ROUTE PRINT command for a host with an IP address or 223.4.5.206.

In Table 5-1, the first route is for certain protocols like DHCP that use 0.0.0.0 as the host address initially. The third network is the local network. The last two routes are the broadcast address for the network and the general broadcast for the 0.0.0.0 network. All of these routes point to the local interface as their gateway.

The other network routes are internal routes that stop at the TCP/IP software itself. This is known as the loopback address and is usually 127.0.0.1 but it just has to start with 127; the rest of the numbers are not significant.

Subnet Broadcast

The broadcast address for every subnet is the network number combined with all the client bits being set to one, or the last address in the network range. The broadcast address is used by many protocols to send a general query to all machines on the network so that one or more of them may respond. For example, the Address Resolution Protocol (ARP) uses a broadcast to the network to locate the IP address of a local host.

Brief Overview of IP Numbering

IP (current version) addressing uses 32 bits conveniently broken into 4 octets, or 4 sets of 8 bits. A Network Number represents one combination of some of those 32 bits, from 2 minimum to 30 maximum. Generally, there are 3 well known groups called Class A, B, and C address ranges. The Class A addresses each have 24 bits that can be used for client numbers or over 16 million IP addresses; the network number in the first 8 bits does not change. The Class B network address is contained in the first 16 bits and the last 16 bits are used for client addresses, allowing for a total of 65,533 values. A Class C network address is in the first 24 bits and the last 8 bits are used for the client addresses.

Example of addresses in each basic class: (using default masks)

Class A 1-126 first octet value (127 reserved for Loopback)

12.33.234.23 255.0.0.0 12.0.0.0 Network 33.234.23 Client

58.38.94.123 255.0.0.0 58.0.0.0 Network 38.94.123 Client

Class B 128-191 first octet value

162.3.24.73 255.255.0.0 162.3.0.0 Network 24.73 Client

144.223.4.5 255.255.0.0 144.223.0.0 Network 4.5 Client

Class C 192- 223 first octet value

212.78.231.59 255.255.255.0 212.78.231.0 Network 59 Client

212.78.231.60 255.255.255.0 212.78.231.0 Network 60 Client

Classes D and E range from 224 to 254 in the first octet and are reserved for special uses.

Broadcast Address

In a Class A network, the 24 bits for the clients represent all the numbers between 1 and 2**24-1. The zero address, like 12.0.0.0, is reserved as the network number. The last address, where all the client bits are set to the value 1 is the broadcast address, 12.255.255.255.

Default Mask

The default mask is a number that identifies all the bits that represent the network number part of the IP address. For a Class A address, 255.0.0.0 is the default mask. This implies that all of the first 8 bits represent the network number. For a Class B address, the first 16 bits indicate the network number; the mask is 255.255.0.0 by default. For a Class C network, the first 24 bits define the network number and the mask is 255.255.255.0 by default.

Subnetting

Subnetting allows you to further break up a standard Class A, B, or C network into smaller sections. By using some of what are normally considered the client bits as part of the network mask, then you can subdivide larger groups into smaller groups of IP addresses.

You can subnet using all but the first bit and last 2 bits of the client addresses. The Class A addresses allow 2 to 22 bits, Class B allow 2 to 14 bits, and Class C allow 2 to 6 bits of the client bits to be used in subnetting the network.

If you use the first 3 bits from any Class A, B, or C network address and include this as part of the subnet, your network masks become:

The following section is a more complete example of subnetting various address classes using 2 bits. If you use the first 2 bits from any Class A, B, or C network address and include this as part of the subnet, your network masks become:

This creates 4 blocks from the original set of addresses for any class since the 2 bits have 4 possible combinations: 00, 01, 10, and 11. Unfortunately, you should not use any block that uses all zeros or all ones. This excludes the 00 and 11 address blocks. You end up with the 01 and 10 blocks or 2 subnets. The formula for the number of groups created is 2**n-2 (2 to the power of n less 2) where n is the number of bits used to create the subnets. This is summarized in Table 5-2.

Network Using Netmask First Client Last Client Broadcast Address

12.0.0.0 255.0.0.0 12.0.0.1 12.255.255.254 12.255.255.255

12.64.0.0 255.192.0.0 12.64.0.1 12.127.255.254 12.127.255.255

12.128.0.0 255.192.0.0 12.128.0.1 12.191.255.254 12.191.255.255

Network Using Netmask First Client Last Client Broadcast Address

162.2.0.0 255.255.0.0 162.2.0.1 162.2.255.254 162.2.255.255

162.2.64.0 255.255.192.0 162.2.64.1 162.2.127.254 162.2.127.255

162.2.128.0 255.255.192.0 162.2.128.1 162.2.191.254 162.2.191.255

Network Using Netmask First Client Last Client Broadcast Address

212.3.6.0 255.255.255.0 212.3.6.1 212.3.6.254 212.3.6.255

212.3.6.64 255.255.255.192 212.3.6.65 212.3.6.126 212.3.6.127

212.3.6.192 255.255.255.192 212.3.6.129 212.3.6.190 212.3.6.191

Network Broadcasts

The network broadcast address is the last address available for a given network; all client bits are set to the value 1. This address is used to send general query packets to all hosts on the network. Broadcast packets are used extensively in all network protocols to establish information about zero or more hosts. The broadcast address for each network in Table 5-26 is the right-most column for each network. Any packet using this address as the destination address forces every host on the network to interpret the request and reply if the request pertains to that host. This is similar to requesting a show of hands from everyone in a theater lineup who has a ticket. This would be considered a broadcast message that everyone listens to, whether or not they respond. For example,, a broadcast message might be to request the hardware address for a given host name.

Local Loopback

There is one specialized network address that is reserved on every machine to represent the "software." This is called the local loopback address. On every machine, the Class A network 127.0.0.0 is used to represent the TCP/IP software. You can ping the loopback to test if your software is running. Normally the loopback is numbered 127.0.0.1 but any numbers after the 127. are acceptable, for example, 127.23.45.67. Just remember the octet numbers vary between 0 and 255.