VoIP Now: Carrying voice over frame relay, IP, and ATM

September 10, 2005

Carrying voice over frame relay, IP, and ATM - Part 2

Voice over Internet Protocol (VoIP) uses IP, which is a connectionless protocol. IP facilitates efficient bandwidth allocation as packets do not follow a preallocated path between endpoints. Paths that are open and are less congested can be used for transmitting the packets. To ensure a high QoS level, it is preferable that the packets get transferred on the same path. Headers used in IP traffic consume bandwidth because of their size, they can be up to 20 bytes, headers in Frame Relays and ATM cells are of 2 bytes and 5 bytes, respectively. The headers contain information that ensures the arrival of the packets at their desired destination as well as in rearranging the packets at the receiver's end.

Fragmentation, jitter buffering, prioritization, voice compression, silence suppression, and echo canceling are some of the methods used in an IP network to increase bandwidth efficiency.

Prioritization: Prioritization is closely linked with QoS. At present, there is no widely accepted QoS standard for IP services. RSVP was an IP QoS protocol under which a sender could try and obtain permission to dispatch his data in a particular manner. It has led to the development of the Differentiated Services Model that uses Type of Service (ToS) to determine the type of traffic at the gateway between the user and the service provider.

Fragmentation: It is carried out to minimize the delay of voice traffic and is performed in a similar manner as in Frame Relay. However, this leads to an increase in the number of IP headers, which means that IP voice traffic may require up to 50% more bandwidth than Frame Relay voice traffic. Improvements in header compression and router technology should help in minimizing bandwidth consumption in IP.

Voice compression: As voice traffic usually travels over links that do not have a very high speed, for example VPNs at many SMBs run only at 28.8 kbps. The ITU G.723.1 standard supports voice compression over IP for dial-up modems and ensures toll quality voice.

Jitter Buffers: These store the packets that arrive so that the delay in the variations is minimized. The setting of the buffer can affect the quality of the conversation. The maximum size of an adaptive jitter buffer can go up to 100 ms to 200 ms. The ideal size is between 30 ms to 50 ms.

Echo Cancellation: Echoes occur because of a mismatch in the impedance in the circuit-switched network or a faulty coupling between the microphone and the speaker of a telephone. VoIP networks can face greater delays than circuit-switched networks and consequently require better echo cancellation techniques. G.165 and G.168 are some of the specs recommended to counter echoes.

Silence Suppression: It is also known as Voice Activity Detection (VAD) and implies the ability to refrain from sending audio packets on an RTP stream during the silent periods, which include the pauses between words and the natural pauses in a conversation. Silence Suppression can help in reducing the bandwidth requirement by 10%.

Voice over ATM: Asynchronous Transfer Mode is an ITU-T standard that lays down the specifications for cell relay of information such as voice, video, etc. The information is relayed in small cells of a fixed size. The technology has the advantage of being high speed and scalable. However, it is an expensive technology. ATM is being increasingly used by corporates to transfer large amounts of voice, graphics, and other such information. ATMs use fixed-size cells that consist of 53 octets/bytes each. The cell consists of a header and a body, with the header consuming 5 bytes and the body taking up the remaining 48 bytes. The small packet size makes ATM suitable for transferring voice and video data as these data types require a steady flow and large data packets take time to download.

Fragmentation: The ATM network uses high-speed switches to run the data through its course. This is possible primarily due to the fragmentation that is built-in into the network. ATM networks use high bandwidths that help in minimizing congestion problems and ensure reliable delivery of data packets. This helps the ATM providers in delivering a high QoS.

Prioritization: VoATM follows the standards laid down in the Adaptation Layer 1 (AAL1) protocol as per the Constant Bit Rate (CBR) service. CBR provides Circuit Emulation Services (CES) that transmits a continuous stream of information; this enables the network to apportion the desired bandwidth to a connection for the transmission period. However, as with circuit switching technology, the voice quality that comes from a regular transmission comes at the price of efficient bandwidth utilization. On occasions, CES can transmit semi-filled cells instead of waiting for the cell to fill. This can lead to wastage of up to 20 bytes of bandwidth per ATM cell. Dynamic Bandwidth Circuit Emulation Service (DBCES) is similar to CES except that it transmits only when the receiver is off the hook.

A Variable Bit Rate (VBR-RT) service as specified by AAL2 is the accepted standard for VoATM. Packets of size 1 to 64 bytes can be transmitted by following the AAL2 standard. These packets are also known as minicells and can be incorporated into an ATM cell. AAL2 supports a variable payload, which helps to improve bandwidth efficiency. AAL2 also supports voice compression and silence suppression and enables multiple voice channels over one ATM connection.

Interoperability between networks: Achieving interoperability between the various networks will allow users to benefit from the best that each network has to offer. ATM offers a high QoS and a good speed, Frame Relay provides an installed base, and IP has a global reach. The extent of compatibility is limited by the prioritization methods and signaling protocols, even though these networks follow similar fragmenting techniques. The level of interoperability will increase with the introduction of standardizations within the protocols, which will facilitate the interworking.

Currently, the Frame Relay Forum has set standards for transmitting voice over Frame Relay; however, there are no standards for voice switching between VFRADs. This has led to the development of proprietary solutions that limits interoperability between the products of different vendors. The use of Switched Virtual Connections (SVC) would entail that paths are defined dynamically, this would increase the scope for interoperability between different solutions.

The interoperability standards for voice and multimedia over IP are defined by ITU H.323. These include endpoint negotiation and the format of the information but not issues such as encoding and security. Also, given the fact that the definitions as given in H.323 can be interpreted in more than one way, a guarantee of interoperability between the products provided by different vendors can not be given. Efforts are underway to provide interoperability between the gateways and gatekeepers provided by different vendors, this is being done by creating an interoperability profile using H.323 and H.225 Annex G standards.

In the absence of standards for these networking technologies, the interworking solutions of the near future will be proprietary. This entails that the users be aware of the technological aspects and that interoperability issues be made transparent to the users. Situations in which an interworking of technologies is desirable include corporate networks that run on Frame Relay and need to communicate with a remote location, their problem can be solved by implementing VoIP, without the need to install a Frame Relay infrastructure. The ability to use multiple voice technology over the same platform also means that migration to another technology need not mean a loss of investment. A new product being developed by RAD will facilitate VoFR - VoIP signaling conversion. This interworking between Frame Relay and IP should be advantageous as the ubiquity of IP services increases.