The scenario for this paper envisages a fast-growing company that has started to experience network routing issues. The company operates a partially meshed WAN that, until recently, has contained only static routes. The company has grown substantially over the past year and the amount of the administrative time necessary to maintain the static routing tables has increased correspondingly. The situation has led to the numerous routing inaccuracies and an increased network downtime as the static routing tables were adjusted in a hurry and often contained mistakes. Based on these considerations, the goal is to implement the dynamic routing protocol, ensuring the shortest path for all network packets and minimizing the routers’ CPU and memory utilization. Other requirements concern the efficient bandwidth use and stable network operation regardless of the certain links’ failures. This paper provides an overview of the available options and indicates the most appropriate solution.
The partially meshed network environment implies that every network has direct links to a number of other networks (Diagram 1), whereas in the fully meshed environment each network is connected to all other networks (Dean, 2009, p. 304). The static routing can provide only fixed paths to the destination network. For example, in order to reach Network E, Router 1 can be configured with the single static route to Router 4 (Diagram 1). In case the link “Router 1 – Router 4” fails, Network E becomes unreachable for Network A.
In contrast to the static routes, dynamic routing protocols can provide adjustable routes to the destination network with priorities based on the number of hops and the links’ state. The routers can use the distance vector protocols, such as RIP (Routing Information Protocol) or IGRP (Interior Gateway Routing Protocol) in order to exchange routing tables that contain all possible paths to every destination network and a number of hops associated with each route (White, 2012, p. 260). However, the regular routing information exchange regardless of any changes to the network topology results in a significant network traffic. If the network topology changes, the timeframe that is necessary for the routers to adapt their routing tables (the network convergence time) is unsatisfactorily long. Moreover, the links’ reliability is not taken into account, as the only basis for the path selection is the number of the network hops (White, 2012, p. 261). These considerations make the distance vector routing protocols inappropriate for the purposes of this scenario.
As opposed to RIP and IGRP, the link-state routing protocols do not rely solely on the network distance. The enhanced version of IGRP (EIGRP) does not have the drawbacks of the distance vector protocols as it initiates the routing information update only when the network topology changes. The routers just send regular “hello” packets in order to check the accessibility of the neighbor networks. However, any topology change can lead to a significant bandwidth consumption. When routers exchange the network updates, they generate an intense traffic that restricts the number of the neighbor routers involved in the process, limiting the EIGRP protocol usability (Dean, 2009, p. 275).
The most appropriate solution for the considered scenario could be achieved using the OSPF (Open Shortest Path First) protocol. It is a link-state dynamic routing protocol, which allows assigning costs to every link depending on the link reliability and connection speed. Thus, the number of hops is relevant but not determinative in the process of the best route selection. If the network is stable, routers do not exchange the routing information, verifying the neighbors’ accessibility by means of keep-alive packets. When the network topology change occurs, routers flood the network by LSA (Link State Advertisement) messages. The routing update process starts immediately on all routers; thus the convergence time on the network running OSPF is much shorter compared to the network with distance vector routing protocols.
Diagram 2 shows a network environment running the OSPF protocol. In order to minimize the routers’ memory and CPU utilization, the networks are grouped into areas. Consequently, routers propagate LSAs only within the OSPF area to which they belong. The fast and reliable network convergence could be achieved by either explicit or implicit LSA acknowledgement. According to Medhi and Ramasamy (2010), “An implicit acknowledgement means that a duplicate of the LSA as an update is sent back to the router from which it has received the update. An explicit acknowledgement means that the receiving router sends a LSA packet on receiving a link state update” (p. 170). In both cases, the router propagates its updated network view to all routers within the same OSPF area.