IP Routing Infrastructure

In order for users and administrators to fully utilize the features of Windows 2000 Server as a router, you need to analyze the network structure and make decisions about what type of routing infrastructure best meets your organization's needs. Table 7.4 describes the various types of routing configurations and their uses.

Table 7.4 Routing Configurations

Static Routed Networks

A static routed IP internetwork does not use routing protocols such as RIP-for-IP or OSPF to communicate routing information between routers. All of the routing information is stored in a routing table on each router. If you decide to implement static routing, ensure that each router has the appropriate routes in its routing table so that traffic can be exchanged between any two endpoints on the IP internetwork.

You can use the network diagram described at the beginning of this chapter to document any static routes in a network infrastructure, and it is an ideal way to keep the routes organized for future reference. Static routes can be entered into the routing table in a Windows 2000 router by using the Routing and Remote Access management console. For more information about adding static routes, see "Unicast IP Routing" in the Microsoft Windows 2000 Server Internetworking Guide.

Before you can use this routing service, you need to configure and enable it from within the management console. For more information about starting and configuring the Windows 2000 Routing and Remote Access service, see Windows 2000 Server online Help. For more information about installing and upgrading Windows 2000 member servers, see "Upgrading and Installing Member Servers" in this book.

You can implement static routes in small networks that require little administration and are not subject to a lot of growth over time, such as a small business with fewer than 10 network segments. However, because they require some administration, you might consider them impractical, especially with the ability of the Windows 2000 Routing and Remote Access service to dynamically build routing information tables for small to large networks using Open Shortest Path First (OSPF) or RIP for IP.

RIP-for-IP Network Design

RIP for IP is a distance-vector routing protocol that dynamically communicates routing information between neighboring routers, automatically adding and removing routes as needed. RIP has a hop limitation of 16. All destinations that are 16 hops and greater are considered unreachable. RIP networks are best implemented in small to medium infrastructures such as medium-sized businesses or branch offices.

Other caveats for using RIP for IP in your network include:

• RIP for IP uses hop count as the metric for the best route. For example, if a site has a T1 link and a satellite backup link, and the costs associated with both of the links are identical, then RIP for IP is free to select either link. To prevent this problem, you can configure the slow link (the satellite) with a cost of two, which forces the router to select the T1 link as the primary link.
• Bandwidth consumption is another consideration because RIP routers announce their lists of reachable networks every 30 seconds. Depending on the size of the network, these announcements can use up expensive WAN bandwidth. Also, as network size increases, the possibility of bottlenecks increases. You can use autostatic RIP updates to reduce bandwidth used by the routing protocol.

Windows 2000 Routing and Remote Access service supports versions 1 and 2 of RIP for IP. RIP version 1 is designed for classful environments and does not announce the subnet mask for each route. If there are routers in your network that only support RIP version 1, and you want to use classless interdomain routing (CIDR) or Variable Length Subnet Mask (VLSM), then upgrade the routers to support RIP version 2, or skip RIP altogether and use OSPF.

You can implement RIP for IP using the following steps:

1. Consult your network diagram to find out where the RIP routers are going to be placed. If you do not have a current diagram, consider designing one before you start. Consider putting routers on a high-bandwidth network in order to keep bottlenecks to a minimum.
2. Determine which IP address scheme is going to be used. Write down which addresses will be used for routers, which ones for servers, and which ones for clients. For example, if you use the private address range of 172.16 0.0/22, you can follow the format shown in Table 7.5.

Table 7.5 IP Address Schemes

Router	Address
Interface on Router1 on the 172.16.4.0/22 network	172.16.4.1
Interface on Router2 on the 172.16.8.0/22 network	172.16.8.1
Domain controller on the 172.16.4.0/22 network	172.16.4.10
Domain controller on the 172.16.8.0/22 network	172.16.8.10
Client on the 172.16.4.0/22 network	172.16.4.20
Client on the 172.16.8.0/22 network	172.16.8.20



3. Next, decide which RIP version is going to be used on each interface. If you are setting up a new network, consider using only RIP version 2, because this version supports CIDR and VLSM. If you have an existing network that uses RIP version 1, consider upgrading to RIP version 2.

OSPF Network Design

RIP for IP is an easy way to integrate a routing protocol into your small- to medium-sized network environment. But, if you have a larger network implemented, RIP for IP might not be sufficient. Another routing protocol that is supported by Windows 2000 Routing and Remote Access is called Open Shortest Path First (OSPF). An OSPF network is best suited for a large infrastructure with more than 50 networks.

OSPF is a link-state routing protocol that calculates routing table entries by constructing a shortest-path tree. It is a more efficient protocol than RIP and does not have the restrictive 16 hop-count problem, which causes data to be dropped after the 16th hop. An OSPF network can have an accumulated path cost of 65,535, which enables you to construct very large networks (within the maximum Time-To-Live value of 255) and assign a wide range of costs. OSPF also supports point-to-point dedicated connections, broadcast networks such as Ethernet, and nonbroadcast networks such as frame relay. One disadvantage to using OSPF is that it is more complex to configure than other routing protocols, such as RIP.

You can structure these networks hierarchically. The sections that follow describe OSPF in more detail.

Autonomous Systems

An autonomous system (AS) is a collection of networks that share a common administrative authority. The following guidelines are recommended when designing an OSPF AS:

• Subdivide the AS into OSPF areas.

  Partition an AS into areas so that OSPF can control traffic to maximize its ability to pass only intra-area traffic, keeping communication to other areas within the AS to a minimum.

• Designate the backbone area as a high-bandwidth network.

  Create a backbone that is capable of maintaining high capacity to help keep inter-area bottlenecks to a minimum.

• Ensure that all inter-area traffic transverses the backbone. Avoid creating virtual links that connect new or changing areas to the backbone.

Figure 7.4 depicts an AS.

Enlarge figure

Figure 7.4 An Autonomous System

OSPF Area Design

OSPF areas are subdivisions of an OSPF AS that contain a contiguous collection of subnets. Areas are administrative boundaries that you can use to separate sites, domains, or groups. Within these areas are networks, which, when joined together through a backbone, form an AS.

In an internal network, configure these areas so that inter-area communication is kept to a minimum. This could include DNS name resolution traffic and Active Directory replication traffic.

One way that traffic leaves and enters an OSPF area is through a router called an area border router (ABR). This router is connected to the backbone called Area 0.0.0.0, which then connects OSPF areas together. ABRs typically have an interface on a backbone area network. However, there are situations where the ABR cannot be physically connected to a backbone network segment. If this happens, you can connect the new OSPF areas to the backbone through a virtual link. Even though this method will work, it is not recommended because it can be complicated to set up and inclined to error. Figure 7.5 shows the backbone, the areas, and a virtual link.

Enlarge figure

Figure 7.5 An OSPF Area Design

To design an OSPF area, follow these guidelines:

• Assign IP addresses in a contiguous manner, allowing them to be summarized. Route summarization is the act of condensing ranges of IP addresses. Ideally, the ABR for an area would summarize all of its network IP addresses into one. This approach condenses routing information, reducing the workload on the ABRs and the number of OSPF routing table entries.
• Create stub areas whenever possible. Keep the following in mind:
  • Stub areas can be configured so that all external routes and routes for destinations outside the OSPF AS are summarized by a single static default route.
  • Any routes that are external to the AS (external routes) cannot be carried by a stub area, including routes that use other routing protocols. This means that stub areas cannot use AS boundary routers (ASBRs).
• Avoid creating virtual links. Virtual links are used to connect new areas in an AS to the backbone. Virtual links can cause routing and other problems, and can be difficult to configure. Always make an effort to connect new areas in your AS directly to the backbone. Ensure this by planning ahead before your AS is implemented.

© 1985-2000 Microsoft Corporation. All rights reserved.