H.323

IP-based networks have become indispensable , thanks to their reliable transmission of data packets. However, they are not optimized for multimedia traffic, which requires careful attention so that packets make it to their destinations on time and in sequence.

If an e-mail message gets pulled apart and reassembled, you'll still wind up with a readable missive. When the packets of a videoconferencing session get jumbled, however, its participants will unwittingly resemble Max Headroom.

H.323, approved by the International Telecommunications Union (ITU) in 1996, addresses this problem. The specification's components define how audio and video should be transmitted, thereby providing real-time applications with the Quality of Service (QoS) they need.

Under The Umbrella

H.323 grew out of the H.320 spec, which was approved in 1990 to support videoconferencing over ISDN and other types of circuit-switched networks. If you've ever used a room-based videoconferencing system, it probably supported H.320.

But because ISDN is still expensive and never took off as many people hoped it would, companies began turning to the much more economical Internet for their communications needs. When that happened , a new protocol was needed to address this inherently unreliable network.

The H.323 standard does not address the network type or Transport layer being used, nor is it dependent on any particular hardware or operating system. The umbrella is big enough to cover plenty of other things, such as specifications for audio and video compressor/decompressor (codec) devices, standards for call setup and control, an interface that supports data conferencing, and provisions for real-time transmissions.

On the audio front, H.323 supports several standards. One of the key components is G.711, a voice standard for digital encoding. Under G.711, voice is usually sent at either 56Kbits/sec or 64Kbits/sec, speeds that can be supported by most network environments. Other audio standards supported within H.323 are G.722, G.723, G.728, and G.729.

As for video, H.323 supports H.261 and H.263, two key video codec standards. H.261 is also used by H.320, H.321 (broadband ISDN, ATM), and H.324 (PSTN analog phone system). H.261 specifies fully encoding some video frames in some instances, and encoding only the changes between a frame and the previous frame in other instances.

The other video codec, H.263, is considered an optional component. It is a newer codec that's backward compatible to H.261 and produces better picture quality.

Important as audio and video standards are, a number of other components within H.323 are also essential. For example, a subgroup usually referred to as call/conference control, or system control, provides the all-important functions of call setup, flow control, and message transmission that oversee the operation of H.323.

The key component in this subgroup is a standard called H.245a control channel protocol that transmits a variety of information through a channel during an H.323 communication. H.245 includes information about flow control, preference requests , and other general commands that need to be sent back and forth during a call. It also defines separate send and receive capabilities and the means to send these details to other devices that support H.323.

Another part of the control subgroup is the Q.931 signaling protocol, which sets up and terminates a connection between two H.323 devices. (Note that while Q.931 uses a fixed IP port, other elements of H.323 use ports that are dynamically allocated. Firewalls must therefore snoop the control channel to determine which dynamic sockets are in use for H.323 sessions, and allow traffic as long as the control channel is active.)

Also within this subgroup is the Registration/Admissions/Status (RAS) channel, a signaling channel that provides a variety of communications among devices in an H.323 call. For example, RAS monitors status, changes in bandwidth, and other functions between H.323 devices and gatekeepers, which I'll look at later on.

Although H.323 is geared toward audio and videoconferencing, it does support data conferencing; the standard supports the ITU's T.120 spec, which defines point-to-point and multipoint data conferencing sessions.

H.323 also includes QoS. It utilizes TCP as the Transport layer for communications that involve dataincluding T.120 and H.245 information. TCP is highly reliable, and the fact that it is prone to delay doesn't affect data information negatively.

But for audio, video, and information going over the RAS channel, H.323 makes use of UDP, which routes packets in sequence and with best effort. When sending this type of information, H.323 supports the Real-Time Transfer Protocol (RTP), an IETF specification for delivering packetized audio and video over the Internet.

Related to RTP and also supported by H.323 is the Real-Time Control Protocol (RTCP), which keeps track of QoS and sends information to the participants about what's happening in the session. Not defined by H.323 is support for RSVP. Currently, only some H.323-compliant products support this method of requesting bandwidth, but this support is expected to increase.

H.323 also supports IP multicast, a way of sending UDP packets to multiple parties without repeating the information being sent out.

Basic Architecture

Building upon the protocols discussed in the previous section, H.323 defines four major network components. See Figure 1 for a sample diagram.

Figure 1: This sample network architecture shows roughly where the four major H323 componentsterminals, gateways, Multipoint Control Units (MCUs), and gatekeeperswould be installed, and how they would communicate with one another and with other network resources.

The H.323 terminal is the LAN endpoint that allows users to communicate with each other in real time. Examples of terminals are videoconferencing or audio conferencing clients, which in many cases are PCs. H.323 stipulates that endpoints must conform to certain standards. They must support G.711 speech compression, H.245 for controlling the media between the clients , the Q.931 signaling protocol, the RAS channel, and RTP/RTCP.

Optionally, terminals can support video codecs, additional audio codecs, and T.120 data conferencing.

Another component of an H.323 network is a gateway, which connects H.323 terminals to other endpoints that do not support the standard generally on circuit-switched networks such as ISDN and the regular phone system. Gateways do this by translating protocols such as audio and video codecs between an H.320 (ISDN) or an H.324 (PSTN) system and clients in an H.323 call, allowing all parties to carry on a meaningful conference. Gateways are also responsible for transferring information between the different networks. Terminals send information to gateways using the H.245 and Q.931 protocols.

Typically, a gateway consists of several parts . One is a switched-circuit network interface, incorporating T1 or ISDN PRI interface cards. Gateways usually also contain NICs for communication with devices on an H.323 network. Other components include digital signal processors, which take care of voice compression and echo cancellation, and a control processor that oversees all other gateway functions.

Gateways are actually an optional component of an H.323 network. If you're communicating directly over a LAN, and not to endpoints on other networks, you won't need a gateway.

On the other hand, if your calls involve three or more endpoints, you need a Multipoint Control Unit (MCU). This consists of a Multipoint Controller (MC), which processes H.245 requests among terminals and controls conference resources, and multipoint processors, which process audio, video, and data.

The fourth and most important piece of any H.323 network is the gatekeeper, which acts as the central point for all calls within its zone. A gatekeeper's zone is defined as the H.323 terminals, translation gateways, and multipoint units over which it has jurisdiction.

Gatekeepers are required to perform four functions. First, they must translate terminal and gateway LAN aliases to IP or IPX addresses. Second, gatekeepers perform bandwidth control, which involves allocating bandwidth during a call; they can refuse to create more connections once a pre-established upper limit for a number of simultaneous conversations has been reached.

A third gatekeeper task is admissions control, which uses RAS messages to authorize network access. The fourth required function is zone management, which involves performing the previous three tasks for all terminals, gateways, and MCUs within its zone.

Gatekeepers can also perform several optional functions. One is call-control signaling, which permits the gatekeeper to process Q.931 signaling messages.

A gatekeeper may also perform bandwidth management, an extension of bandwidth control, which means it can determine when there is no available bandwidth for a call, or if there is no more available bandwidth when a call in progress requests more.

Other optional gatekeeper services include call authorization, which involves the acceptance or rejection of calls based on criteria such as time of day, type of service, and lack of bandwidth. Gatekeepers also may perform call management, which involves keeping track of H.323 calls in progress to know which terminals are busy. This helps gatekeepers redirect calls or save call-setup time by not trying to reach a terminal already in use.

The Next Step

In January 1998, the ITU ratified the second version of H.323, which includes a number of enhancements as well as several new features.

Major enhancements include the ability for endpoints to set QoS through RSVP; gatekeeper redundancy; support of URL-style addresses; the ability for gateways to let a gatekeeper know of its present resource availability; and increased audio and video capabilities.

The second version of H.323 also includes faster call setup and new security features. Supporting a standard called H.235, this latest version offers authentication, integrity for data packets, privacy, and nonrepudiation.

Since 1997, the number of products supporting H.323 has increased dramatically. Not only do desktop videoconferencing products support it, but so do IP telephony gateways and whiteboarding products. And, as is the case whenever a critical mass of a particular type of product hits the market, interoperability tests have been performed on H.323-compliant products, making the decision on what to buy that much easier.

While most of us have yet to participate in our first videoconferencing session or place our first telephone call over IP, standards such as those defined within H.323 will make it easier for vendors to create products that work well with others, and that deliver real-time video, audio, and data to a much wider audience.

Resources

For a primer on H.323, see www.databeam.com/h323/h323primer.html.

For an in-depth look at the standards efforts behind H.323, visit the ITU's Web site at www.itu.ch/.

To see more detailed information about H.323 gatekeepers, gateways, and standards, visit the H.323 center at www.elemedia.com.

This tutorial, number 126, by Anita Karv, was originally published in the January 1999 issue of Network Magazine.