Home > Articles > Operating Systems, Server > Microsoft Servers

Windows Server Reference Guide

Hosted by

IP Routing for Subnets

Last updated Sep 26, 2003.

Only the smallest networks today consist of a single TCP/IP subnet. Networks that consist of more than one subnet must have some mechanism for forwarding IP traffic to the proper destination network. For this reason, you have to perform some planning of IP routing at the subnet level—in other words, prior to forwarding packets onto IP routers on the broader Internet. IP routing uses default gateways on a subnet, whose addresses represent routers, to forward traffic to the proper destination subnet. These routers can be either hardware devices or routing software installed on a PC.

The Concept of the Routing Table

Every IP host is responsible for the job of routing to some degree, of forwarding IP messages on to new locations. To that end, every host maintains a local routing table. You don't see it too often, but Windows Server does make it available for you to look at. Here in this Virtual Server 2005 R2 environment, for example, is a routing table for a system that's connected to two other hosts through separate interfaces, labeled "Local Area Connection 2" and "Local Area Connection 3." ("Local Area Connection 1" died a painful, though brief, death.) Obviously, a routing table shows everyplace that can be considered "there," or in other words, is a list of all the "there's" in a network you can get to from "here." Now, there's a few default ways to loop back from here to here (127.0.0.1), of course, but most important are the connections between "here" and the computer at 192.168.1.105 for LAC 2, and between "here" and 192.168.0.101 for LAC 3.

Figure 1

Figure 1 An example of a routing table for a Windows Server 2003-based IP host on a Virtual Server 2005 setup, maintaining two network connections.

A static route is precisely what it sounds like: a point-to-point connection in terms of IP addresses, between "here" and "there." "There," in this case, is a specified interface, such as "Local Area Connection"—what Windows otherwise classifies as "network connections." Essentially, all network connections lead out. A dynamic routing system enables Windows to determine which is the best way that leads "out;" though for a small network, that way might always be the same way anyway. In the network whose routing table is pictured here, it would be dangerous to establish a static route, because obviously both of the systems "there" are obtaining their IP addresses via DHCP. Which means they're subject to change.

When a system or a router isn't using a static route, then it's using routing protocols such as Routing Information Protocol (RIP) and Open Shortest Path First (OSPF) to dynamically learn the routes through the network. Windows Server supports both static and dynamic routing tables.

Dynamic routers use routing protocols to update routing tables by exchanging information about known routes. Most dynamic routers use one of three routing protocols: Routing Information Protocol (RIP), Open Shortest Path First (OSPF), or Enhanced Interior Gateway Routing Protocol (EIGRP). Windows Server 2003 includes standard support for the first two, perhaps because the name of the third one alone was just too long. Of these three, Windows Server 2008 only supports RIP; the removal of OSPF support there has been noted by Microsoft but not explained.

Setting Up a Dynamic Routing Protocol

To add active support for dynamic routing protocol, from the Routing and Remote Access MMC snap-in, in the left pane, choose the name of your domain controller. For Windows Server 2008, choose between IPv4 and IPv6. Then right-click General, and from the popup, select New Routing Protocol.

From the dialog box, choose RIP or OSPF (in WS2K3), then click OK. It'll be installed first, and in a moment, you'll see it in the middle pane. To set it up, you right-click on the new protocol in the left pane, and from the popup, select Properties.

RIP is the easiest dynamic routing protocol to set up. It's a distance vector protocol that uses hop counts plus a cost to calculate the best path to take to a destination network. Both Windows Server 2003 and 2008 support version 2, whose added capabilities include multicast routing update, simple password authentication, and route filtering. However, they do support incoming packets based on RIP version 1. Filters can be applied to keep certain routers from learning about particular subnets and routes to reach those subnets.

OSPF is a link-state routing protocol. Instead of broadcasting or multicasting its routing table, OSPF sends other routers the status of its links. These links are then added to a link state database on each router. This database provides an efficient map of the network's topology. This makes OSPF more scalable and responsive to change than either RIP 1 or 2. Its scalability and responsiveness comes from the low overhead required to pass updates, even in very large networks. However, its scalability comes at the expense of greater complexity in planning, configuring, and administering OSPF router ports.

As the size of the link-state database increases, the load on system resources increases as well. To facilitate scalability, OSPF divides the network into areas. Each area is connected to other areas using a backbone. Routers maintain link states only for their own area and for the backbones. This effectively segments the database and the network into a manageable size. Routers directly connected to the backbones are known as area border routers (ABRs).

While RIP is the easiest of the dynamic routing protocols, since every RIP router must maintain a routing table for the whole network, it loses efficiency as the network's size grows. The availability of areas in OSPF segments the network and decreases the overall size of the routing tables that must be maintained. OSPF's link-state database system also leads to faster adaptation to changes in network topology, while guaranteeing that routes are loop-free.

References

Books and E-books