Almost all the network traffic in the world these days is Transmission Control Protocol/Internet Protocol (TCP/IP) over Ethernet. From a practical standpoint, this means that if you are just starting to learn about networking, you should focus on this technology. However, broadcasters may also encounter User Datagram Protocol (UDP)/IP over Ethernet. This introductory tutorial on networking software will focus on must know items for broadcast engineers.
Fundamental principles
It is important to understand the fundamental principles that affect the behavior of networks within and between broadcast facilities:
- Ethernet is a packetized networkAll Ethernet data, whether it is an automation log file, a video clip or a live video stream, is sent in packets. This means that the data (sometimes called the payload) is cut up into pieces and loaded into packets for transport across the network. For example, think of packets as tractor-trailer rigs; the payload is loaded into the trailer. At the receiver, the chunks are reassembled in the order in which they were sent.
- Packets have a source address and a destination addressAddress information is contained in a header, which can be quickly read by anyone who needs to know where the packet came from or where it is going. Imagine that the tractor-trailer has two large, easily-read labels on the cab, which say where it started from and where it is headed. (See Figure 1.)
- Packets are self-routingSwitches and routers read the addresses in each packet to figure out where it needs to go. This is contrary to the routed video model where input from an external control system (a router control panel) determines the A/V path through the router.
- Packets travel independentlyOnce a packet is launched onto the network, there is no way to know what path it will take to get to the destination (unless the network is small). The specific route may change for a number of reasons, including equipment fault, congestion or maintenance. Each packet is routed independently of all other packets. This can lead to unexpected behavior. For instance, packets launched in succession may arrive at the destination out of order; that is, if you launch two packets, the second packet may arrive before the first packet since it may be routed across a shorter path. Also, the transit time across the network may vary from packet to packet.
- Networks lose informationNetworks drop packets. In fact, in some cases, networks drop packets on purpose! Packets may be lost because of equipment failures, but they may also be intentionally dropped when the network becomes congested.
- IP over Ethernet is not a guaranteed transfer mechanismUnderlying network mechanisms do not guarantee that a launched packet will reach its destination. At the lower level, Ethernet is designed as a “launch and forget” proposition. It is up to higher software systems to deal with delivery issues. UDP does not make any attempt to recognize and correct for lost packets. TCP attempts to resend lost packets, but if the errors are too great, TCP will give up entirely.
- Network file transfers are an “all or nothing” affairWhen a file is transferred over a network, the transfer either succeeds or it fails. If a single bit is corrupted during the transfer (and TCP is unable to correct it by resending the data), then the entire transfer will be aborted. Network protocols are designed in this way to ensure that errors are not propagated across the network.
- Networks are designed using layered protocolsA long time ago, network designers figured out that writing one huge networking application was a bad idea. What if the hardware changed from wire to fiber? What if people needed more error protection? What if chip technology changed to allow faster network speeds? In one huge application, the entire thing would have to be rewritten every time one of these things changed.Instead, network designers came up with a plan for a layered system; thus, you may hear that a router operates at Layer 3, etc. Different technologies may be substituted within a specific layer. You can have TCP/IP over Ethernet, or you can have UDP/IP over Ethernet. TCP and UDP operate at the same layer, so you can change between TCP and UDP without having to make changes to IP or Ethernet.
- Broadcasters have complex networking needsOn the one hand, broadcasters are just like any other business — they need networking technology to move files, pay bills, send e-mail, print documents and do all the other things businesses do day in and day out.On the other hand, broadcasters have a tendency to stress their networks, particularly technical networks. Increasingly, broadcasters are moving large video files and transmitting professional quality streams across their networks both within and between facilities. Broadcasters need to realize that core networking technology was never designed with these applications in mind. If you are working with a network used to support operations, normal business networking rules apply. If you are working with a video production network where users are exchanging A/V files and high-resolution streams, then things may not work as expected, and you may have to employ some tricks to get your network to behave reliably. In some cases, you may have to move away from TCP/IP over Ethernet entirely.
- QoS is important for broadcastersAs you can imagine, some of the fundamental characteristics of networks mentioned above do not work out well for broadcasters in a professional video environment. Fortunately, the needs of the broadcaster are lining up well with the consumer market where being able to handle video is becoming a priority. This means that QoS capabilities are being built into more equipment over time. Using QoS, you can guarantee the level of performance across a network. But you should know that QoS is not a thing; you cannot buy QoS. You also cannot set the QoS switch on your router and get guaranteed delivery.Instead, QoS is a framework of policies and technologies that can be used to control the quality of service to be delivered by your network. At a high level, QoS works by controlling traffic admitted to the network, marking traffic by type (voice, video, data, etc.), and establishing priorities for traffic types so that lower priority traffic gets dropped first when the network gets busy. (See Figure 2.)QoS can establish guaranteed routes for particular flows so that all packets in a stream flow from source to destination via the same path, no matter what. This avoids out-of-order packets and packets with wildly varying arrival times at the destination.
- All networks are not the sameThere are several broad classifications of networks, and it is important to know the differences between them. There are local area networks (LANs) and wide area networks (WANs). A LAN is usually confined to a single facility. There are WANs that run between facilities over a wide geographical area. There is the public Internet, and there are private managed networks. Both the public Internet and private managed networks are WANs. (They may be comprised of several WANS.) However, in the case of private managed networks, QoS is carefully controlled to give users predictable performance. There is no guarantee on the public Internet.
There are virtual private networks (VPNs) as well. VPNs connect more than one facility together so that all computers look like they are in the same facility, even if they are physically located miles apart.
Finally, there are virtual LANs (VLANs), which segment network traffic on the same physical network. You might put all of the traffic on one VLAN, all of the news department on another one and the sales department on a third. All three VLANs are physically connected to the same routers and switches in the facility. To the clients on the network, these computers appear to be running on three physically separate networks.
Brad Gilmer is president of Gilmer & Associates and executive director of the Advanced Media Workflow Association.
Send questions and comments to:brad.gilmer@penton.com
