[ Pobierz całość w formacie PDF ]
.Ostensibly, LSAs are transmitted because a router detects a change in the state of a link(s).Therefore,after receiving an LSA of any type, an OSPF router must check the contents of that LSA against theappropriate portion of its own routing database.This can't be done until after the router uses the new datato form a new perspective of the network, which is done via the SPF algorithm.The result of this outputis the router's new perspective of the network.These results are compared with the existing OSPF routingdatabase to see whether any of its routes have been affected by the network's change in state.If one or more existing routes must change as a result of the state change, the router builds a new routingdatabase using the new information.Duplicate LSAsGiven that LSAs are flooded throughout an OSPF area, it is possible that multiple occurrences, known asinstances, of the same LSA type will exist simultaneously.The stability of an OSPF network, therefore,requires a router to be able to identify the most current instance of the duplicated LSA.A router thatreceives two or more instances of the same LSA type examines the LS Age, LS Sequence Number, andthe LS Checksum fields in the LSA headers.Only the information contained in the newest LSA isaccepted and subjected to the processes described in the preceding section.http://wwwin.cisco.com/cpress/cc/td/cpress/fund/iprf/ip2912.htm (14 of 20) [02/02/2001 11.39.24]OSPFThe Link-State Acknowledgment PacketThe fifth type of OSPF packet is the link-state acknowledgment packet.OSPF features a reliabledistribution of LSA packets (remember that LSA stands for link-state advertisement, not link-stateacknowledgement).Reliability means that receipt of the packet must be acknowledged; otherwise, thesource node would have no way of knowing whether the LSA actually reached its intended destination.Therefore, some mechanism was needed to acknowledge receipt of LSAs.This mechanism is thelink-state acknowledgment packet.The link-state acknowledgment packet uniquely identifies the LSA packet of which it is acknowledgingreceipt.This identification is based on the header information contained in the LSA's header, includingLS sequence number and advertising router.There needn't be a one-to-one correlation between LSAs andacknowledgement packets.Multiple LSAs can be acknowledged with a single acknowledgment packet.Calculating RoutesOSPF, despite its complexity, calculates the costs of a route in one of two remarkably simple ways:A non-bandwidth-sensitive default value can be used for each OSPF interface.OSPF can automatically calculate the cost of using individual router interfaces.Regardless of which method is employed, the cost of any given route is calculated by summing the costsof all interfaces encountered along that route.A record is kept of the summed costs to known destinationin OSPF's shortest-path tree.Using AutocalculationOSPF can automatically calculate the cost of an interface.This algorithm is based on the amount ofbandwidth that each interface type supports.The sum of the calculated values of all interfaces in a givenroute forms the basis for OSPF routing decisions.These values enable OSPF to calculate routes based, ata minimum, on the bandwidth available per link in redundant routes.Figure 12-9 presents a samplenetwork to demonstrate this point.Figure 12-9: Autocalculated costs of the links.http://wwwin.cisco.com/cpress/cc/td/cpress/fund/iprf/ip2912.htm (15 of 20) [02/02/2001 11.39.24]OSPFIn Figure 12-9, the cost of the WAN route between a host in network 193.1.3.0 and an end system innetwork 193.1.4.0 is 138.This cost is the sum of the two T1 links between those networks, each witha cost of 64, plus the cost of the Ethernet interface to network 193.1.4.The cost of the Ethernetinterfaces at the origination and destination points are not included in the OSPF cost calculation becausethe OSPF calculates only the costs of outbound router interfaces.Table 12-1 summarizes the automatically calculated costs for each of the interfaces used in Figure 12-9'snetwork diagram.Table 12-1: Calculated Costs Per Interface TypeInterface Type Calculated Cost100-Mbps FDDI 110-Mbps Ethernet 101.544-Mbps T1 serial link 6456-kbps serial link 1,768Using Default Route CostsIt is usually in your best interest to have OSPF automatically calculate route costs, although this may notbe possible.Older routers, for example, might not support the autocalculation feature.In such cases, allinterfaces will have the same OSPF cost.Therefore, a T3 will have exactly the same cost as a 56-kbpsleased line.Clearly, these two facilities offer very different levels of performance.This disparity shouldform the basis of informed routing decisions.There are, however, circumstances that may make the use of default route costs acceptable.If yournetwork consists of relatively homogeneous transmission facilities, for example, default values would beacceptable.Alternatively, you can manually change the cost metrics for specific interfaces.This wouldenable you to shape traffic patterns in your OSPF network as you see fit, while still using predominantlyhttp://wwwin.cisco.com/cpress/cc/td/cpress/fund/iprf/ip2912.htm (16 of 20) [02/02/2001 11.39.24]OSPFdefault routing costs.Homogeneous NetworksIn a homogeneous network, all the transmission facilities are the same.All the LAN interfaces would be10-Mbps Ethernet, for example, and all the serial WAN interfaces would be T1s.In such a scenario,using the default values would not likely cause routing problems.This would be particularly true if therewere little, if any, route redundancy.To illustrate this point, consider the network diagram in Figure 12-10.Figure 12-10: Acceptable use of OSPF's default interface values.In Figure 12-10, a default value of 1,768 was assigned to each of the interfaces.All the WAN links,however, are T1s.Given that they are all the same, it doesn't matter whether the value assigned them is 1,128, 1,768, or 1,000,000! Routing decisions, in a homogeneous network, become a simple matter ofcounting and comparing hops (albeit in multiples of the interface costs).This would be true regardless ofhow much, or how little, route redundancy existed in the network [ Pobierz całość w formacie PDF ]