Background: Know: Recognize:
Anticipates: Fault management
When analysing the network layer from the perspective of NOC, it is important to identify the impact that routing metrics used by different routing protocols will have on the overall performance of the network. One might be tempted to think that because routing protocols automate the determination of the “shortest” path then this is one less aspect for the network operator to configure and/or monitor. But this naive approach ignores not only the fact that the network operator is responsible for specifying the link costs that the routing protocol will use to perform its calculations, but also the impact the selection of the routing protocol will have on fault management. First, the link cost specified should account for the monetary cost of running the link, its performance and the enforcement of company policies:
Monetary cost: Primarily defined based on the cost in dollars of using each path. For example, the company could have access to different paths provided by different ISPs which charge a different rate for their use.
Performance: Link performance metrics such as the bandwidth, number of hops, packet loss, latency (delay), path reliability, throughput and load will each have a different weight determined by the network operator according to the type of data to be carried.
Policies enforcement: Defined by the network operator in order to control the access to certain paths which could be dedicated for governmental, contractual and/or security reasons.
Second, when focusing on fault management we need to identify the two different main types of routing protocols: the ones based on link-state (such as OSPF and IS-IS) and the ones based on a distance vector (such as RIP and IGRP). Routers using a distance vector protocol do not have knowledge of the entire path to a destination, they only know the interface to which a packet should be forwarded and the distance from its destination because each router has a routing table only for the physically connected neighbour routers. On the other hand, link-state protocols make every node construct a connectivity map of the network from which to obtain the best path to store in the routing table. When comparing both types of routing protocols we can see that by having information about multiple paths to a destination to coose from (link-state) instead of only the connected neighbours (distance vector) the router is ready to route traffic on another path when needed without computation delay, making link-state protocols better suited to manage faults.