All 14 articles in this series are now available in our free 62 page eBook ‘Understanding IP Broadcast Production Networks’ – download it HERE.

All articles are also available individually:

Routing protocols allow distributed routers to communicate with each other so that IP packets can be sent to different subnets in the most efficient way possible. One of the fundamental concepts of distance vector routing is that each router only has knowledge of the neighboring routers it’s attached to. This greatly simplifies the network administration as each router does not need an understanding of the whole network topology.

Figure one shows a simple distributed studio network connecting cameras to a production switcher and monitors from different geographical dispersed layer-2 networks. Three routers are used to demonstrate the distributive approach of networks and resilience. For example, the production switcher has two distinct paths back to the cameras; link A-B, and link B-C to A-C. Deciding which path to take and then what happens in the event of a link failure is the subject of this chapter.

Distance vector routing (DVR) is a protocol which allows routers to decide which is the most optimal route to take to reach the destination host. Intuitively, we might think that if a datagram was sent form camera-1 to the rack monitors we should use link A-C, that is router A then C. But from the metrics we can see that route A-B and B-C has measure 4 (3 for A-B, and 1 for B-C), and route A-C has measure 5. In this example, the most optimum route is A-B then B-C.

The measures are calculated using several different protocols; routing information protocol (RIP), and interior gateway routing protocol (IGRP) being another. RIP is one of the oldest DVR protocols and uses hop counting as its measure. In diagram one, if RIP was used instead of IGRP, then a measure of one exists between router A and B, a measure of one between routers B and C, and a measure of one between routers C and A. So, to travel from router A to C, we have a measure of two for A-B then B-C, and one for A-C. This implies A-C is the fastest route but as RIP only counts hops and does not take into consideration any propagation delays between the hops, the measure may not be accurate. Consequently, IGRP improves on RIP by adding node and propagation delays to the measure. Link B-C might be fiber and A-C might be satellite, so B-C is clearly faster and has a lower metric.

When the network is first switched on, each routers’ DVR is zero and has no knowledge of any of its neighboring routers. As each router initializes it looks at its routing table and advertises the subnets it can see to its neighbors. Each router continues to periodically advertise the subnets it can see to its neighboring routers. As the routers build up knowledge of each other’s routes, and build up knowledge of the measures, they are able to select the best route to send a packet.

If camera-1 sends video to the racks monitors, it will first break the video into small IP datagrams. The IP source address will be 10.0.1.0 and the IP destination address will be 10.0.3.1. Router A will have determined the most efficient route to send the datagram will be along link A-B. The IP source and destination addresses stay the same, but the Ethernet headers will have a source address of port 3 on router A and a designation address of port 4 on router B.

When router B receives the datagram it knows, from the metrics in the routing table, populated by the distance routing vector algorithm, that the most efficient route to 10.0.3.1 is via router C, therefore it will keep the IP source and destination addresses the same, but change the Ethernet source address to the MAC of router B port 3, and the destination MAC to port 4 on router C.

If link A-B was to fail, then router A would eventually realize that there was no traffic or control messages coming from router B, and also any packets sent by router C to either router A or the production swithcer would time out. Router C would remove the route on port 4 from its routing table and send error messages to its neighbors advising network timeouts on link A-B. From router A’s point of view the next best metric to the racks monitor is 5, the link A-C. It would automatically start sending its camera-1 packets along link A-C. The network administrator would be alerted to the lost error messages reported by router C and take action to fix the link.

As networks increase in complexity the measure process can take time to reach its optimum routing, this is referred to as convergence.

There are some problems with distance vector routing especially with infinite loops. The count-to-infinity problem is one of these and suffers from the fact that new routes are advertised quickly, but problems are detected slowly, and the routers become unstable because a route that is being advertised as valid, is in fact failing.

Although DVR’s use minimal resource and require basic administrator knowledge, they do have some other disadvantages namely the entire routing table is sent every 30 to 90 seconds taking up valuable bandwidth.

Link state protocol is the newest edition to routing protocols and overcomes some of these limitations by only sending changes in routing tables, and updates are triggered by events such as a subnet failing, instead of being sent periodically resulting in reduction of bandwidth used. They also allow routers to gain a greater knowledge of the whole topology of the network, increasing network efficiency and resilience. However, they are extremely complicated and need advanced administrator knowledge to make them work and keep them maintained.