BY LINDEN DeCARMO

[Go to a glossary for this article.]

The cable industry is agonizing over whether its telephony clients should be fat or thin. Distributed Call Signaling (DCS) proponents feel that the smarts should be placed in an intelligent -- or fat -- client. By contrast, Network Call Signaling (NCS) supporters believe that all intelligence should be placed in network servers.

Network Centric
A DCS network contains two core elements: intelligent endpoints and network helpers. DCS endpoints are typically either a cable modem or a Multimedia Terminal Adapter (MTA). Their primary responsibility is translating packet-based audio streams into a format understandable by conventional phones.

A DCS endpoint is dramatically different from a simple analog phone. Not only is it likely to have a powerful microprocessor, but it also contains a plethora of programmable software services (features) such as call forwarding and voice mail. Besides these software features, DCS endpoints are responsible for both call signaling and resource management.

Call signaling refers to the messages that are exchanged between endpoints when a call is being connected. Because DCS is an extension of the Internet Engineering Task Force's (IETF) Session Initiation Protocol (SIP), it is compatible with emerging industry call signaling standards.

Resource management is necessary to ensure the smooth flow of audio content during a multimedia session. Typically, the calling and destination endpoints negotiate the data rate at which they will be streaming and reserve the appropriate network bandwidth for this data. If these resources are not reserved, random traffic may cause choppy audio or jerky video.

Network Helpers
Although a DCS endpoint is intelligent, it is ignorant of network resources or the location of other endpoints. Thus, it must rely on helpers to navigate the cable telephony network.

For instance, the endpoint obtains packet routing services from a Cable Modem Terminal Server (CMTS).

Another helper is the Call Management Server (CMS). The CMS provides an endpoint with bandwidth management, authentication, and privacy features. The first responsibility of the CMS is to authenticate the person using the endpoint and authorize service.

Figure 1. DCS architectural diagram.

Authentication involves verifying that the endpoint is legitimate (i.e. the user's bill is paid and the endpoint is using a valid IP address). Once it is authorized, the CMS screens the features that can be executed on the endpoint. For instance, the endpoint may contain the software for call waiting. However, the user may not have paid for that service. In this scenario, the CMS would not permit call waiting on that endpoint.

A second function provided by a CMS is bandwidth management. Each CMTS contains virtual "bandwidth gates" that control the flow of multimedia packets between endpoints. Once both endpoints are authorized, the CMS instructs the bandwidth gates about the types of content that may be exchanged between endpoints. For instance, the bandwidth gates may be authorized by the CMS to permit voice calls but refuse video calls.

The final duty of the CMS is to ensure privacy. Recipients can interrogate conventional SIP packets and retrieve the IP address of the sender. Hackers may attempt to exploit this loophole to wreak havoc in a network. Consequently, the CMS scrambles the packets so that both the calling and receiving parties are secure from hacking.

Network Scalability Is Essential
Since each endpoint maintains call status, CMS's are responsible only for translation, authentication and privacy transactions. Therefore, the designers of DCS feel that it will scale better on larger networks since each additional endpoint provides negligible increases in the responsibilities of the CMS.

Another theoretical strength of DCS is its ability to perform selective service upgrades without affecting the entire network. For instance, a limited number of endpoints can be used to test new features or services. If the new feature fails, only the test endpoints are affected.

The Push For Simplicity
As voice over IP (VoIP) evolves, protocols are modified to address the limitations in their predecessors. For example, H.323 was the first standardized VoIP protocol. Unfortunately, it is a complex protocol that is vague in crucial areas. These weaknesses limit H.323's usefulness in consumer telephony devices.

VoIP vendors realized they needed a simpler, more robust protocol in order to penetrate consumer markets. Consequently, the concept of a Device Control Protocol (DCP) was created. DCPs remove all the call signaling, resource management, authentication, and service execution responsibilities from the endpoint and focus only on the minimum features necessary to enable telephony. This reduction in complexity should result in more stable endpoints.

Over the past year, several DCPs have been refined by the IETF to form the Media Gateway Control Protocol (MGCP). PacketCable, the organization charted by CableLabs to establish cable telephony standards, has sanitized MGCP for cable usage and folded it into Network Call Signaling. However, NCS is more than a simple DCP. It defines the billing, authorization and security features that are required in a cable telephony environment.

Unlike DCS, NCS endpoints (commonly referred to as Residential Gateways or RGWs) contain minimal intelligence. They are responsible only for detecting events (such as off-hook or a key press), reporting these events to a server and transporting media streams to or from the destination endpoint. Each RGW connects to a Media Gateway Controller (MGC). MGCs are sometimes referred to as Call Agents when they act as agents on behalf of the RGW to complete calls and provide custom calling features.

The MGC is a superset of a DCS CMS. Not only does it perform translation, authentication, and security services, but it is also responsible for executing features on behalf of the RGW. Thus, services such as call forwarding are implemented on the server rather than the endpoint as in DCS.

The Fruits of Simplicity
MGCP's simplicity should enable Residential Gateway vendors to dramatically improve quality and decrease development time. Furthermore, selective service upgrades are possible with NCS by placing the new services on a dedicated MGC specifically for beta testing. Test RGWs connect with the MGC that is running the new service. If problems arise, the damage is limited to a few RGWs and a test MGC.

Figure 2. NCS selective service rollout.

Another significant advantage to NCS is that all endpoints can be instantly upgraded by updating software on the MGC. To illustrate, once the call forwarding feature is added to a MGC, all RGWs that attach to that MGC can automatically take advantage of the call forwarding functionality.

Which is Better?
As we discovered, DCS is less burdensome to servers and therefore may potentially scale better on large networks. Unfortunately, the industry's H.323 experience indicates that intelligent endpoints are difficult to build and even harder to interoperate. Furthermore, intelligent devices in consumers' homes are likely to require considerably more tech support hand-holding should hardware or software problems arise.

By contrast, NCS's simplicity encourages interoperability, stability, and superior service features. However, it places significantly more strain on the network server and requires more intelligence in order to scale correctly.

Neither solution is architecturally superior and both involve tradeoffs. Currently, there are more vendors claiming NCS and MGCP compatibility, but deployment is in the very early stages and this ratio may change as cable telephony networks mature. It is likely that DCS and NCS will continue to evolve and co-exist for a significant period of time before one is declared a winner.

Linden deCarmo is a senior software engineer at NetSpeak Corporation where he develops advanced call agent software. He is the author of technical articles and a book entitled Core Java Media Framework. You can reach him at [email protected].

NetSpeak is a leading developer of Advanced Intelligent Network (AIN) technologies for Internet Protocol (IP) telephony. NetSpeak solutions enable today's market leaders to build next generation communications networks by providing the market's premier intelligent software products that enable solutions for concurrent, real-time interactive voice, video and data communications over packet networks. NetSpeak's products provide service providers and enterprises with the flexibility they need to cost-effectively add advanced communications capabilities - enabling an easy transition into the rapidly growing Internet telephony market.