All 14 articles in this series are now available in our free eBook ‘Understanding IP Production Networks - 2026 Edition’ – download it HERE.

All articles are also available individually:

Routers and switches reduce traffic congestion by sending packets and frames only to the devices that require the data and not spraying the whole network with them. As well as reducing congestion this also helps improve security as administrators can limit which hosts can exchange data; IP-cameras, IP-production switchers, and computers etc.

An IP packet consists of two parts, the header, and the payload. In the case of video over IP, the camera will break its video stream into smaller chunks until they fit into the payload of an IP packet. The header consists of the source IP address, that is the address of the device sending the packet such as a camera, and the destination address of where the camera wants to send its video stream to, for example a production switcher.

If we assume that the network we are using in the studio is CAT6/Ethernet, the camera will have its own unique media access control (MAC) address, and the production switcher will also have its own MAC address. These are unique Ethernet addresses that are hard coded into the devices during manufacture and are different for every Ethernet enabled unit that leaves the production line.

Routers & Switches

A network often consists of routers and switches where switches distribute traffic within the locality of the connected devices, and routers route packets between localized networks. For example, studio-1 consisting of four cameras and a production switcher will only distribute data within studio-1. However, if camera-1 from studio-1 needs to be sent to studio-2, then a router is required.

Ethernet networks are defined as a network which share the same Ethernet broadcast address. It is possible to have multiple switches cascaded across many studios but whenever an Ethernet broadcast message is sent, for example, when using an address resolution protocol (ARP) protocol, then every device in every studio will receive this message, thus creating unnecessary traffic and potentially increased latency. To stop this from happening, routers are often used to separate the networks, so the broadcast traffic is only sent to the devices within the locality of a studio, thus reducing congestion, network saturation, and packet loss.

VLANs also solve this problem and are discussed in a later chapter.

In Figure-2 we have a simple studio network where a production switcher is connected to three cameras via two layer-3 routers. It is possible to connect all these devices together on one layer-2 switch or one layer-3 router but that would leave us with a single point of failure and probably over-load the switch or router due to the high bandwidths video-over-IP demands.

In this example, it’s reasonably academic as to whether we use routers or switches. Switches were originally favored over routers for localized networks as they were much faster in their datagram processing. As the switch doesn’t have to change values such as the Time To Live (TTL) counter, then they were traditionally much faster at processing. It’s only when we get into much larger and complex networks do we need to start separating layer-2 switching and layer-3 routing. Layer-2 networks offer flexibility with VLANs, but layer-3 designs provide better scalability.

The router uses a combination of the IP address and netmask, called the Network-ID to route packets to other routers. A typical network ID in the class-less system would be 10.1.1.0/24. The “24” refers to the subnet mask and the 10.1.1.0 is the network. In this instance the router would route any packet in the address range 10.1.1.0 to 10.1.1.255, as it only looks at the first 24 bits of the IP address, and each number within the dot is 8 bits. As another example, a network ID such as 10.1.1.0/8 will contain all hosts with an IP address range from 10.0.0.0 to 10.255.255.255. Subnets are often aligned with Ethernet networks to aid administration and security.

Netmasks are a method to speed up routing as the router does not have to compare the whole IP address to send the packet to the correct port. Instead, it is only concerned with the subnet comparison. With the 8-bit network ID, only the first 8-bits of the IP address need to be compared against the netmask, thus greatly reducing computation time.

Static routers work on a next-hop system, that is each router only knows of the existence of subnets connected to its ports. This simplifies the network design and keeps the router database to a manageable size, but it does suffer from variable and unpredictable latency when using routing protocols such as Equal-Cost Multi-Path (ECMP), so great care must be taken when using IT industry type routing.

In Figure-2, Ethernet is being used as the physical interface and Camera-1 will have to set the source and destination MAC addresses of the Ethernet frame leaving camera and entering the switch port. A MAC address is a unique number issued by the IEEE for every piece of equipment that is manufactured. When the IP packet leaves camera-1 it is encapsulated by the MAC addresses of Camera-1 and Port-1 of Router-A.

An address resolution protocol (ARP) is instigated by camera-1 which broadcasts an Ethernet message asking “who has IP address 10.0.3.1 (the production switcher)? And what is your Ethernet address?”. The broadcasting of ARP messages is restricted to local subnets and routers to stop the entire network being flooded with ARP query messages.

When the production switcher is connected to Port-2 of router-B, the ARP will be instigated, and router-B will add to its routing database the MAC and IP address of the production switcher connected to its Port-2. When router-A generates an ARP request asking “who has IP address 10.0.3.1”, router-B will respond on its Port-1 and router-A will then know that the production switcher is connected to its Port-1 and update its IP routing tables accordingly. It’s important to note that router-A does not know that it is connected to router-B, it only receives an ARP response saying that the production switcher is available on this link.

This method ensures that each router only needs to keep a record of the next hop for routing and is also the reason why TTLs exist in the IP header; an incorrectly configured network could create a loop that allows an IP packet to traverse forever around the network. For example, in Figure-2, if port-3 on router-B is connected to port-10 on router-A, then a loop will occur, thus causing massive congestion. At each router, the TTL value is decremented and if it reaches zero (as would happen with an incorrectly configured network), the packet would be deleted, thus stopping congestion.

At this point router-A answers the ARP query from camera-1 with the address of the MAC address for port-1. Camera-1 sets its own MAC address as the source MAC address, and the destination address is the MAC address of port-1 on router-A. Port 5 on router-A will have a different MAC address than ports 1 to 3 as it’s connected to a different network segment.

Each router builds a database of the connected devices so will substitute the destination MAC addresses as required. The routers in Figure-2 provide an IP network using Ethernet layer-2 as its transport stream, so, if the studio technical director wants to send camera-1 from studio-1 to studio-2, then a router with the appropriate forwarding tables will need to be provided. This stops unnecessary traffic from studio-1 flooding studio-2, and vice versa.

Throughout the whole process of sending a packet from camera-1 to the production switcher via router-A and router-B, the source and destination IP addresses did not change at all. However, the source and destination Ethernet MAC addresses changed at each network link.

It’s entirely possible that the cameras on router-A were at the Superbowl stadium, with router-B and the production switcher at the studio in New York, with an IP satellite link connecting them. The source and destination IP addresses did not change, but the MAC and physical connectivity mapping automatically changed as the frames moved between devices. From an IP point of view, we do not know or care how the packet travelled between routers A and B as the underlying complexity of the physical, electrical and optical routing was abstracted away from us. And this is one of the greatest strengths of IP.

Dangers Of Asynchronous Networks

IP networks are designed to be asynchronous as this facilitates packet switching which greatly enhances flexibility, scalability and efficient use of bandwidth. Consequently, packets can bunch together to the point where they saturate a link and there is no space left to insert new packets, leading to dropped packets. Switches and routers partially deal with this in enterprise designs by adding egress and/or ingress buffers to the inputs and outputs of the device allowing the packets to be temporarily held in the buffer until space becomes available on the link.

A blocking switch or router is typical of those found in datacenters as the statistical nature of the packet distribution in the temporal domain of transient delivery devices such as databases, websites and streaming servers, means that space often appears to allow insertion of new packets. If too many packets are input to the buffer, or the output packets are not removed from it fast enough, the switch or router buffer overflows, leading to packets being simply dropped.

This is not an error in the design, but a consequence of blocking devices working in asynchronous networks as dropped packets will cause the TCP/IP sender to reduce their sending datarate, but at the expense of variable and indeterminate latency. TCP/IP is the dominant protocol in enterprise datacenters, but broadcasters prefer UDP/IP as latency is much more predictable.

Although this may be acceptable in enterprise networks, even with only occasional congestion and packet loss, this method of operation is totally unacceptable for broadcast television where dropped packets can have a significant impact on the viewing experience for both video and audio. Furthermore, the potential for dropped packets due to buffer overflow is exasperated by the switch or router design as the backplane bandwidth will probably not be able to simultaneously route all inputs to all outputs at full bandwidth. Again, this is by design as enterprise datacenters will assume this full link bandwidth situation rarely occurs.

Non-blocking switches guarantee that all the switch or router input and output ports can simultaneously route all their packets without loss at full bandwidth. In other words, their backplane and buffer design facilitates full bandwidth data exchange throughout the device. This is something broadcast engineers have taken for granted with SDI/AES switching matrix switchers due to the circuit switched nature of their designs as the matrix switcher has full backplane bandwidth.

However, the same cannot be assumed with IP network switches and routers. The cost of a non-blocking switch or router compared to the blocking type is orders of magnitude more expensive, therefore, great care must be taken when using blocking switches to guarantee the devices do not over subscribe the ports and buffers and saturate the links, as packet loss is virtually guaranteed to occur. Non-blocking switchers and routers are preferable, but their cost may inhibit their use.