PDF of this tutorialSometimes referred to as a media gateway, the switching/bearer transport platform is hardware that sits at the edge of a network and takes in a packet and/or circuit containing voice or data traffic and switches it to a voice or data network. Media gateways come in many different flavors depending on the breadth of definition. The most popular consist of Class 4 and Class 5 replacement functionality and a voice over digital subscriber line (VoDSL) gateway. Media gateways are part of the physical transport layer and are controlled by a call control engine or softswitch (also called a media gateway controller), which provides instructions to direct voice traffic. Media gateways are at the heart of the transformation of the voice network, as they are essential to migrating voice traffic onto a packetized network. As part of packetizing voice traffic, a media gateway adapts (by using compression and echo cancellation) the packetized traffic, creates and attaches an IP header and/or ATM header, and sends the packet through the network according to instructions provided by the softswitch.
While a media gateway can be physically located almost anywhere within the network, depending on the network architecture and the features it is intended to support, all media gateways share certain features including the following:
- Scalability—A media gateway needs to be able to scale to support hundreds of thousands of telephone calls (called DS–0s, running at 64 Kbps per line) to parallel the scalability of the existing PSTN switches.
- Support for several types of access networks—Needed support includes wireless, fiber, cable, and copper. In addition to electrical interfaces, a media gateway needs to support a variety of optical interfaces (including OC–3, OC–12, OC–48, and OC–192 speeds).
- Carrier-class reliability—Also known as five nines (99.999 percent) reliability (i.e., less than five minutes of downtime per year) and network equipment building standards (NEBS) certification (the Telcordia quality rating for meeting environmental stress tests), reliability is extremely important to service providers because it enables them to fulfill customer contracts. Most carriers cite reliability as the impetus to transform their current architecture.
- Interworking functionality—Media gateways are capable of supporting multiple voice and data interface protocols and compatibility between them by converting circuit traffic to packet traffic and vice versa.
- Interoperability—Most networks are a compilation of multivendor solutions, making interoperability essential for success.
- Control support—To enable communication between the media gateway and a softswitch. The most common languages (or protocols) emerging for communication between these devices are MGCP and Megaco.
- Switching—A media gateway must handle switching and media processing, based on an ATM, IP, or TDM switching fabric.
- Voice transportation—There are 3 transport standards used for transporting voice traffic: TDM (traditional circuit-switch method), ATM AAL–1/AAL–2, and IP–based RTP/RTCP (over ATM or pure–IP transport).
A packetized approach to transmitting voice faces a number of technical challenges that spring from the real-time or interactive nature of the voice traffic. Some of the challenges that need to be addressed include the following:
- Echo—Echo is a phenomenon where a transmitted voice signal gets reflected back due to unavoidable impedance mismatch and four-wire/two-wire conversion between the telephone handset and the communication network. Echo can, depending on the severity, disrupt the normal flow of conversation and its severity depends on the round-trip time delay if a round-trip time delay is more than 30 ms the echo becomes significant making normal conversation difficult.
- End-to-end delay—Voice traffic is most sensitive to delay and is mildly sensitive to variations in delay (jitter). It is critical that end-to-end delay is minimized to hold interactive communications. Delay can interfere with the dynamics of voice communication, in the absence of noticeable echo, whereas in the presence of noticeable echo, increasing delay makes echo effects worse. When delay reaches above 30 ms, echo canceller circuits are required to control the echo.
- Packetization delay (or cell construction delay)—Packetization delay is the time taken to fill in a complete packet/cell before it is transmitted. Normal G.711 pulse code modulation (PCM) encoded voice samples arrive at the rate of 64 Kbps, which means it can take approximately 6 ms to fill the entire 48-byte payload of an ATM cell. The problem can be addressed either with partially filled cells or by multiplexing several voice calls into a single ATM virtual circuit channel (VCC).
- Buffering delay—Sometimes, due to delay in transit, some cells might arrive late. If this happens the ATM segmentation and reassembling (SAR) function provided by the adaptation layer might have to under run with no voice data to process which results in gaps in conversation. To prevent this, the receiving SAR function would accumulate a buffer of information before starting the reconstruction. In order to ensure that no under runs occur the buffer size should exceed the maximum predicted delay. The size of the buffer translates into delay, as each cell must progress through the buffer on arrival at the emulated circuit's line rate. This implies that the cell delay variation (CDV) has to be controlled within the ATM network.
- Silence suppression—Voice, by its nature, is variable. In fact, a typical conversation has a speech activity factor of about 42 percent due to pauses between sentences and words where there is no speech in either direction. Also, voice communication is half-duplex, which means that one person is silent while the other speaks. One can take advantage of these two characteristics to save bandwidth by halting the transmission of cells during these silent periods. This is known as silence suppression.
- Compression algorithms—G.726 adaptive differential pulse code modulation (ADPCM) and G.729 adaptive code excited linear prediction (ACELP) are the two major compression algorithms that are used. The benefit of compression is efficient use of bandwidth. Most voice packets are transmitted today using G.711 encoding that does no compression and therefore adds further delay.
