Video networks: Packaging and shipping data

We've always packaged, or wrapped, video with other information before sending it on its way. Originally, we wrapped it with sync pulses. Now we convert analog picture elements into sequences of numbers (data), but we still surround them with timing information. Instead of sync pulses, we now add data strings called “start of active video” (SAV) and “end of active video” (EAV). We can also add audio and other ancillary data now in SMPTE259M/CCIR601 bit streams. We can use these streams to move data other than baseband video and audio signals. The Serial Digital Transmission Interface (SDTI) replaces the baseband video payload with what is usually MPEG data.

As technology progressed, analog video and audio came to live most of its life as digital data, and we began to move it in ways developed by the computer and telco folks. That is, we developed sets of rules — protocols — to break the data into “chunks.” Until very recently, the three most common protocols we used for this purpose have been Frame Relay, Internet Protocol (IP) and Asynchronous Transfer Mode (ATM). All three of these protocols use virtual, connectionless paths. This means that there is no physically switched, dedicated path between end locations. The data from one location is merged — via time multiplexing — with other traffic headed the same direction.

Data traffic can be very “bursty,” and many data users require some guarantee that the data will arrive at its destination in a timely manner. To achieve this, frame relay uses paths called permanent virtual circuits (PVCs) to connect end users. Many other data sessions can share different links of the path, but the data carrier sets up the virtual paths to be there all the time, whether or not the user has data to send. If no data is sent, the time slots (or, more precisely, the frames) devoted to that data customer are empty.

What can be confusing about these protocols is that some are often wrapped within others. For example, IP frames often can be inserted as the data payload into ATM cells. In another example, a user's data can be wrapped in an error-correction scheme called Transport Control Protocol (TCP). This TCP data can then be wrapped a second time in IP, which provides destination addressing. Additionally, IP can be wrapped in Ethernet protocol for transport over the user's local network.

Different protocols are used in different situations according to the advantage they offer. The advantage of frame relay is that its frame lengths are long, generally 128 bytes and up. This means that, for lower-bit-rate (bandwidth) paths, there is less overhead. The trade-off is that data requiring fast access to the network has to wait longer for its turn. To solve this and other problems, ATM has shorter cell lengths. IP is much cheaper to implement than ATM. But ATM is good for carrying time-sensitive material only in networks that are not congested, or through congested networks where quality-of-service levels along the path need to be tunneled out. IP traffic outside the carrier's synchronous optical network (SONET) backbone is routed via routers, which are cheaper than switches. Routers are generally software-driven devices, and thus the propagation delay through them is longer than the hardware-oriented switches used for ATM. Another protocol, User Datagram Protocol (UDP), can be used in place of TCP since, unlike TCP, it doesn't require re-sending cells that are lost. The re-sending of lost cells and the wait to assemble the re-sent cells in the proper order would greatly hamper the high-bit-rate/real-time nature of television bit streams.

Another protocol and set of hardware now used to transport television data is a specification known as P1394. As with Fibre Channel, P1394 carries SCSI commands known as Serial Bus Protocol. P1394 automatically negotiates itself into a “tree” network with nodes establishing parent/child relationships. P1394 divides its time between asynchronous “normal” data transfer and isochronous “ship and pray” data transfer from one node to any other node interested in receiving the data. Isochronous data transfer is used for moving video between P1394 devices. These architectures provide common access to a file system. They use special switches — content-aware packet switches — that transport both isochronous and asynchronous data, extending the network's reach via optical fiber.

There are two types of network topologies used to attach storage to data networks.

Storage Area Networks (SANs) and Network Attached Storage (NAS) comprise one or more storage devices connected to clients as servers, usually via Fibre Channel. While the SAN is based on Fibre Channel connectivity, the NAS is based on Ethernet connected to a “protocol converter” that converts NAS to a SAN. Compared to a SAN, NAS storage greatly simplifies client access to storage.

Since SONET is usually the backbone that moves data over WANs, you need to gain access to the SONET carrier. This is usually achieved by using the Incumbent Local Exchange Carrier (ILEC) — often the local phone company — or a Competitive Local Exchange Carrier (CLEC) to provide connectivity from your facility to the ATM carrier's point of presence (POP). Once your data is at the POP, the ATM carrier charges a port charge. This is a subscription into the SONET network. One DS-3 bit stream is inserted into an STS-1 frame, which in turn is inserted into an optical carrier (OC-1) stream. An OC-1 stream is 51.840Mb/s with about 8Mb/s overhead on top of the 44.21 DS-3 stream. OC3 (155Mb/s) carries three DS-3 streams, or 84 DS-1 streams, or 2016 DS-0 (voice) streams. It is possible to achieve a concatenated mapping, by which all the bandwidth is given to a single user (no digital hierarchy at all) such as an ATM user. (A small “c” indicates a concatenated frame, e.g.: STS-1c.) OC-12 carries 12 DS-3 streams (622Mb/s). OC-48 carries an STM-16 bit stream, which handles 48 DS-3 streams. OC-192, which can carry 192 DS-3 streams, is currently the highest bit rate available. On a SONET ring, you often find different kinds of traffic. Some carriers have voice, IP and ATM traffic all traveling over the same OC ring. Although the different traffic can be thought of as separate virtual networks, they all travel over the same physical fiber ring in STS frames.

As you can see, there are many ways to wrap and ship video data over networks. Different protocols and network topologies have been developed to solve particular data-transport needs, and they often work in concert to move the data from place to place.

Jim Boston is director of emerging technology for the Evers Group.