Dedicated bandwidth over cable: Simplifying the migration to VoIP service

Over the past seven years, cable operators have spent $50 billion on network upgrades, driving fiber deeper into the network and increasing bandwidth capacity of their hybrid fiber/coax (HFC) networks. As revenues from residential video and cable modem subscribers have leveled off due to market saturation and competitive pressures, cable companies have looked toward offering additional services to bundle with existing services. IP telephony has long been considered a key feature to a bundled cable offering; however, two of the key barriers in wider scale deployment have been the capability of HFC access technology and the ability to transition to softswitch architecture.

For an HFC access technology to be a feasible means of voice service delivery, it must meet a wide range of service and operational requirements. First, the technology must support quality-of-service (QOS) levels suitable for supporting real-time applications such as telephony that have strict latency, jitter and packet loss requirements. Given that many cable operators have made huge investments in legacy equipment, such as Class 5 voice switches, it is crucial that an access technology provide a migration path that initially leverages and maximizes return on existing equipment.

The access technology

Dedicated IP is a packet-based technology for HFC access networks that provides dedicated, deterministic, scalable bandwidth of up to 40 Mb/s to each subscriber over point-to-point logical connections over existing HFC plant. Dedicated IP is an in-band RF technology based on quadrature amplitude modulation (QAM) and quadrature phase-shift keying transmission in within the existing cable spectrum. Dedicated IP was conceived to allow cable operators to deliver high revenue-generating IP-based services such as IP telephony to both residential and commercial subscribers. The dedicated IP architecture coexists with existing technologies on the cable plant, including analog video, digital video and cable modem systems, and requires no changes to the HFC network.

In dedicated IP, bandwidth is allocated to each subscriber in a deterministic manner using a round-robin serving scheduling mechanism within each frequency channel based on downstream time division multiplexed (TDM) sub-channels and upstream time division multiple access (TDMA) sub-channels illustrated in Figure 1 below. Each subscriber is allocated one or more TDM channels in the downstream providing 5 to 40 Mb/s service scaleable in 5 Mb/s increments, and one or more TDMA channels in the upstream providing 500 Kb/s to 8 Mb/s service scaleable in 500 Kb/s. Because TDMA and TDM channels are not oversubscribed to multiple subscribers, the dedicated IP architecture completely eliminates subscriber-based contention on the access link thereby inherently providing low latency, low jitter connectivity to support voice applications.

 

Figure 1: Dedicated IP channels

Dedicated IP is supported by two key network elements: a headend access router (HAR) deployed at the headend or distribution hub and an access gateway (AGW) deployed at the customer premise. The HAR provides standard packet network interfaces such as Gigabit Ethernet to the WAN and F-type interfaces to the HFC combining and splitting networks. The HAR behaves as an IP router supporting IP forwarding between network interfaces to the WAN or MAN and subscriber-facing dedicated IP channels over the HFC access network. The AGW provides an F-type interface to the HFC network and standard LAN interfaces such as 10/100Base-T Ethernet to customer premise equipment such as a subscriber telephony gateway (STG). The AGW may provide one or more RJ-11 analog voice ports that directly interface analog handsets.

Because dedicated IP is purely an access technology for HFC, the operator is not constrained to supporting any particular VoIP architecture for primary or secondary line voice. Consequently, dedicated IP could be used to support VoIP architectures based on any combination of H.323, session initiation protocol (SIP), media gateway control protocol (MGCP) or Megaco. Nonetheless, since one of the prevalent goals of the cable industry is to provide a VoIP migration strategy for cable operators with substantial investments in legacy circuit-switched telephony equipment, this article will overview both a near-term and longer-term solution for IP telephony over dedicated IP toward that end. 

Hybrid IP-TDM architecture

The hybrid IP-TDM architecture is primarily based on MGCP and a voice access gateway (VAG) that provides an interface between the public network and a VoIP network running over the HFC access network. MGCP, which is defined in the Internet Engineering Task Force (IETF) RFC 2705, is based on a distributed packet telephony system centered on two types of devices: a media gateway controller (MGC), which is often referred to as a call agent, and media gateway (MG). MGCP specifies how MGs are controlled by MGCs assumed to reside in a different network location from the media gateways they control.

The hybrid IP-TDM architecture is based in part on the components below, which are illustrated in the Figure 2. Note that operations support system (OSS) servers used for provisioning, billing and customer care are not shown. 

• Voice access gateway (VAG) consisting of a partial media gateway (MG), signaling gateway (SG) and media gateway controller (MGC)

• Local digital switch (Class 5 switch)

• Telcordia GR-303 interface between the Class 5 switch and the VAG

• Subscriber telephony gateway (STG)

Figure 2: Hybrid IP-TDM telephony architecture using dedicated IP for access

The MG component of the VAG interfaces the HFC access network via a standard networking interface such as Gigabit Ethernet and interfaces a local digital switch (LDS) via a Telcordia GR-303 or V5.2 interface. The VAG converts TDM signaling, received over the GR-303 interface from the Integrated Digital Terminal (IDT) on the LDS, to MGCP messages sent over dedicated IP channels to the subscriber telephony gateway. The VAG may also encode TDM signaling using real-time transport protocol (RTP) as described in RFC 2833 to minimize delays in ring signaling. The MCG component of the VAG controls connection state, event detection and line signaling of remote STG devices at the customer premise via MGCP. 

The LDS interfaces to the public network and provides line side call features to the packet network via a GR-303 interface. In the VAG architecture, all CLASS and regulatory features are supported by the LDS, which also provides most all of the progress tones and can also detect DTMF signaling.

The STG, deployed at the customer premise, provides one or more RJ-11 analog telephony interfaces to the VoIP network. The STG may either be a stand-alone unit that interfaces the access gateway via a 10/100Base-T Ethernet connection or integrated into the access gateway. The STG consists of a subscriber line interface circuit (SLIC), coder decoder (CODEC), tone generator and detector and microcontroller. The SLIC provides the analog interface to a subscriber's telephone and the CODEC converts analog voice signals to a digital format and vice versa. The tone generator and detector is used to generate progress tones; however, the tone detector may be bypassed if the VAG encodes telephony tones and signals into RTP payloads transmitted to the STG as per RFC 2833. In the reverse direction, the STG encodes DTMF digits into an RTP stream and sent to the VAG where the packets are converted to audio and sent toward the switch over an assigned timeslot. 

There are several reasons why the hybrid IP-TDM approach may be attractive to the cable operator for initial deployments. First, using a VAG allows the operator to leverage existing Class 5 switches. Because Class 5 switches support more than 1500 call features, not immediately migrating to a pure software call feature implementation mitigates some risk for the operator until the extensive set of call features can be validated over time in field trials. Also, many of the emergency and operator services tandem switches support only multifrequency (MF) signaling and not Signaling System 7 (SS7) signaling, which is currently supported by most of the softswitch vendors.  

Pure VoIP architecture based on a softswitch

The hybrid IP-TDM telephony architecture described above is presented as an intermediate step between circuit-switched telephony over cable and the softswitch-based approach illustrated in Figure 3. The softswitch architecture separates the signaling and call control path from the media path unlike in the hybrid IP-TDM approach where the VAG translates the call signaling and media at the same point in the network. The softswitch model is similar to that of circuit-switched telephony where the SS7 network supports call control signaling over a separate network path than the bearer path for the call itself.

Figure 3: Softswitch architecture using dedicated IP for access

The softswitch architecture centers on MGCP or Megaco, which provides distributed media gateway control, and an IP call signaling protocol such as SIP. H.323 could also be used as the IP call signaling protocol; however, SIP is often favored in part because SIP messaging is much simpler than that of H.323, and SIP interoperates well with MGCP/Megaco because each protocol uses messaging based on the session description protocol (SDP). SDP, which is defined in RFC 2327, provides a format for describing session information and defines the syntax of the actual messages contained in MGCP and SIP payloads.  

In this scenario, the media gateway controller, also known as a softswitch for which the architecture is named, provides all of the calling features such as Call Waiting and Caller ID as well as emergency 911 services. Note the distinction in this approach as opposed to the VAG-based model where the Class 5 public network switch supports such call features.  

In the softswitch architecture, the media gateway controller performs signaling translation, and media gateways provide packet to circuit-switched media translation between the VoIP network and the public network. In the SIP/MGCP model, the MGC translates between SIP and SS7 messaging such as IDSN User Part (ISUP), which is used to set up and tear down phone calls between public network switches.   

Figure 3 illustrates two possible modes of providing telephony services to subscribers. The first approach is similar to that of the VAG-based model in which analog telephones connect via RJ-11 connections to a subscriber telephony gateway, which is either a separate device behind the access gateway or an integrated access gateway module. In this case, the subscriber telephony gateway is controlled by the media gateway controller via MGCP as before. Because the softswitch architecture is purely messaged based, it imposes additional tone detection and generation requirements on the STG. For instance, the STG must generate all progress tones (dial tone, etc.) and detect dual tone multifrequency (DTMF) as well as dial pulse signals, unlike the VAG model in which telephony tones and signals may be transported using RTP.

The diagram above also illustrates SIP-based telephony support for SIP user agents such as the "IP phone," which is shown connected the access gateway via Ethernet. SIP is based on a client-server model where calls originate at the client and terminate at a server. In the network diagram, the SIP proxy server behaves as a server terminating a SIP session to the IP phone and a client supporting a SIP session to the MGC acting on behalf of the IP phone and providing for call completion to the public network. Not shown in the illustration are other types of SIP entities such as SIP redirect servers, user agent servers and registrar.

As in the VAG model, the switch router provides for IP transport and is essentially transparent to the overlying VoIP architecture beyond that of providing adequate quality of service for voice. Note that the switch router architecture is extensible to supporting an integrated media gateway for directly interfacing the public network.

Conclusion

Dedicated IP allows cable operators to deliver voice services that cannot be easily provisioned with legacy shared bandwidth cable modems and provides a migration path from TDM-based to IP-based telephony services. Cable operators can also deliver video and data services over dedicated IP without disrupting legacy cable services. Bundling voice with video and data offers the opportunity to substantially increase revenue per customer and reduce churn.

Further, dedicated IP allows cable operators provide voice, video and data services to high revenue generating small to medium enterprise (SME) subscribers. Because many businesses are located near existing cable plant, providing a coax drop to the business location is relatively inexpensive.

Cable IP telephony over the dedicated IP architecture helps fulfill the promise of advanced services that enable cable operators to offer differentiated service offerings to subscribers. Not only is IP voice highly scalable, lowering the cost of delivering voice services, it also reinforces the opportunity to provide business-class services that are immediate revenue generators.

Ryan Leatherbury is Director of Systems Architecture for Advent Networks.

Visit Advent Networks online.