3.1.3 Architecture Overview

The basic architectural components for videoconferencing/Voice over IP are illustrated in Figure . This diagram is protocol-independent and emphasizes the similarities between H.323 and SIP architectures.

Figure: Basic Videoconferencing / Voice over IP Architecture

Two independently administered and geographically separated domains are depicted. Each domain contains endpoints, an authentication server, a person directory, a call server, and one or more gateways and routers. Some components, such as endpoints and call servers, are commonly deployed; other components such as policy servers are rarely deployed at this point in time.

For both H.323 and SIP architectures, users have a device such as a mobile phone, PC or laptop with microphone and camera, or other "appliance". All these devices contain the client software for the protocol. The H.323 protocol refers to these clients as "Endpoints"; the SIP protocol refers to these clients as "User Agents". These clients are able to directly place calls to other clients of the same protocol. The client software also encodes and decodes the multimedia streams.

Both H.323 and SIP architectures utilize a call server. Call signaling managed by these call servers may require that the endpoint registers at the server. The H.323 protocol refers to this server as the "Gatekeeper" while the SIP protocol refers to "Proxy Servers". Call servers are used for signaling information outside of the media stream, including user authentication, managing usage, user's location, etc.

The Multipoint Control Unit (MCU) is used for multi-party conferencing. In H.323, the MCU is treated as a special endpoint; it's role is to receive incoming streams from each conference participant and select one of these streams to send back to each conference participant except the one selected for this broadcast. SIP MCU's are in early beta testing at the time this cookbook was written.

Gateways serve as protocol translators; to place a call from an H.323 endpoint to a SIP user agent, an intervening gateway would be needed to translate both the call setup/session initiation steps. Both H.323 and SIP use RTP to encode the media.