The Best Damn Windows Server 2003 Book Period- P81

The Best Damn Windows Server 2003 Book Period- P81:The latest incarnation of Microsoft’s server product,Windows Server 2003, brings manynew features and improvements that make the network administrator’s job easier.Thischapter will briefly summarize what’s new in 2003 and introduce you to the four membersof the Windows Server 2003 family: the Web Edition, the Standard Edition, theEnterprise Edition, and the Datacenter Edition.
Nội dung trích xuất từ tài liệu:
The Best Damn Windows Server 2003 Book Period- P81766 Chapter 22 • Planning, Implementing, and Maintaining a Routing Strategy There are several shortcomings to RIP version 1: I RIP version 1 uses MAC-level broadcasting, requiring all hosts on a network to process all packets. I RIP version 1 doesn’t support sending a subnet address with the route announcement. This can be a problem when there is a shortage of available IP addresses. I Because RIP version 1 route announcements are being addressed to the IP subnet and MAC-level broadcast, non-RIP hosts may also be receiving the RIP announcements, con- tributing to the broadcast clutter and possibly lowering the efficiency and performance of your network. I By default, every 30 seconds, RIP routers broadcast lists of networks they can reach to every other adjacent router. Again, this can contribute to lower network performance. I RIP version 1 does not handle subnetted addresses well, since it doesn’t send the subnet address along with the broadcast. I RIP version 1 provides no defense from a rogue router. A rogue router is an RIP router that advertises false or erroneous route information. I RIP version 1 is difficult to troubleshoot. In general, most problems in RIP routing stem from incorrect configuration or from the propagation of bad routing information. So, what does RIP version 2 do to attempt to correct the problems with RIP version 1? I RIP version 2 advertisements include the subnet mask with the network ID. I RIP version 2 sends multicast announcements to the multicast IP address 224.0.0.9 with a time to live (TTL) of 1 instead of broadcasting announcements, so it does not require IGMP. I RIP version 2 allows for authentication to substantiate the source of the incoming routing announcements. I RIP version 2 is compatible with RIP version 1. RIP routers begin with a basically empty routing table and start sending out announcements to the networks to which they’re connected.These announcements include the appropriate routes listed for all interfaces in the router’s routing table.The router also sends out a RIP General Request message asking for information from any router receiving the message.These announcements can be broadcast or multicast. Other routers on other networks hear these announcements and add the original router and its information to their own routing tables.They then respond to the new router’s request for information.The new router hears the announcements from these other routers on the network and adds them and their information to its own routing table. After the initial setup, the RIP router will send out information based on its routing table.The default time period is 30 seconds. Over time, the routers of the network develop a consensus of what the network looks like.The process of developing this consensual perspective of the network’s topology is known as convergence. Basically, this means that the network’s routers individually agree on what the network looks like as a group. It is this very process of convergence, however, that can Planning, Implementing, and Maintaining a Routing Strategy • Chapter 22 767sometimes lead to problems. A typical network using convergence is shown in Figure 22.3. One ofthe occasional problems that occurs is called counting to infinity. Let’s look at how that happens.Figure 22.3 Typical Network Using Convergence Router D Router B Router C Router A In our example, we will assume that Router A has failed. With its failure, all the hosts on the Anetwork will no longer be accessible from the other three networks. After missing six updates fromRouter A, Router B will invalidate its B–A route and advertise its unavailability. Routers C and Dremain ignorant of the failure of Router A until notified by Router B. At this point, both Router Band Router D still think they can get to Router A through Router C, and they raise the metric ofthis route accordingly. So, Routers B and D send their next updates to Router C. Router C, havingtimed out its route to Router A, still thinks it has access through Router B or Router D.Thus, aloop is formed between Routers B, C, and D, based on the mistaken belief that both Routers B andC can still access Router A. With each iteration of updates, the metrics are incremented an extrahop for each route.This count speeds up the process by which the router approaches its definitionof infinity—the point where the router says the destination is unreachable. There are two methods of preventing this counting to infinity loop: split horizon and triggeredupdates. If the router is implementing split horizon, routes will not be announced back over theinterfaces by which they were learned.The limitation of the split-horizon approach is that a routewill not timeout until it has been unreachable for six tries, so each router has five opportunities totransmit incorrect information to the neighboring routers. If the router is implementing splithorizon with poison reverse, routes learned on interfaces are announced back as unreachable. Splithorizon with poison reverse is much more dependable than simple split horizon. However, althoughsplit horizon with poison reverse will stop loops in small networks, loops are still possible on larger,multipath networks. Fault toler ...