The basic purpose of a broadband network is to satisfy the increasing demand for bandwidth-intensive integrated services. Services are offered through the use of both narrowband and broadband connections operating simultaneously over high-speed Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH) and dense wavelength-division multiplexing (DWDM) transmission gear, controlled by high-capacity packet-switching equipment.
Asynchronous Transfer Mode (ATM)-based networks integrate circuit-mode (voice) digital-switching methods with packet-mode (data) switching. The crucial difference between these two methods is that in circuit mode, separate service channels require a dedicated amount of bandwidth--64 kbits/sec--for each voice connection. This bandwidth cannot be shared or used by any other application until the connection is broken. In contrast, ATM networks operating in the packet-mode allow for the multiplexing of different virtual channels (VCs) consisting of separate service channels on a transmission link. Bandwidth potentially can be shared between applications on the core network because ATM switching equipment will not transmit information if there is no data being sent by the source. By limiting the amount of idle channel information being sent within its cells, and allowing the multiplexing of various traffic sources (voice, data, or video) over the same link, ATM enhances the carrier's flexibility to efficiently manage network bandwidth.
From a carrier perspective, there are several industry motivations for transporting voice over packet (VOP). One is the opportunity for additional revenue by transporting voice across the spare capacity of data networks. Savings are achieved by integrating the voice and data traffic across a common backbone. A VOP network promotes bandwidth efficiencies. It also reduces capital expenditures, the number of network elements, and the size of the operations and support staff.
But how can providers migrate to a VOP network without compromising the expected reliability in the interexchange carriers' (IXCs') voice networks? The availability of the signaling network should be at least 99.999%, allowing for only five minutes of downtime a year. This availability should be achieved while keeping the architecture simple and the implementation flexible enough to incorporate developing standards.
An IXC voice network is typically based on dedicated connections between two endpoints. It consists of a class 3 switch connected to a class 4 (tandem) or class 5 (end-office) switch using Signaling System 7 (SS7) as the call-processing protocol (see Fig. 1). Enhanced services such as enhanced 8XX and calling cards are provided using a service control point (SCP) and/or a service node (SN).
The backbone is designed to optimally transport DS-0 (64-kbit/sec) and T1 (1.544-Mbit/sec) circuits-the units of interchange between the traditional voice switches. Large carriers derive the bulk of their revenues from DS-0 and T1 services.
Traditional voice networks groom time-division multiplexing (TDM) circuits at a granularity of DS-3 (45 Mbits/sec) using SONET/SDH add/drop multiplexers (ADMs) and digital crossconnects (DCS) as necessary. The SONET/SDH ring architecture is used for restoration.
The deficiencies in this voice transport model stem mostly from the segregation of the voice and data traffic. Statistical multiplexing gains are not realized because the architecture uses dedicated voice connections, versus the virtual connections utilized in the data call model. Separation of the voice and data traffic also introduces a higher network build-out cost and as a result, more support staff with different expertise are necessary. Similar services are offered for voice and data customers, but by separating the physical infrastructure, duplication of effort occurs in the service development process.
From a transport perspective, SONET/SDH ADMs and DCSs are expensive solutions to scale as traffic demands grow. With data-centric networks, it's no longer obvious that the job of restoration should be performed by SONET/SDH. Data services that utilize the Internet protocol (IP) at layer 3 can often tolerate interruptions or degradations in the link between the user and server. Updates in the IP routes are performed by the layer 3 routing protocol--open shortest path first (OSPF), border gateway protocol (BGP), or routing information protocol (RIP)--to ensure restoration. In many cases, layer 2 restoration schemes can recover service from a problem. For example, rerouting of ATM permanent virtual circuits (PVCs) or virtual paths (VPs) can restore the connections between the IP routers. Layer 1 restoration at the SONET/SDH level should be available if desired by the customer, but it shouldn't be the only option.
Successful network design provides the maximum amount of bandwidth flexibility while minimizing the cost of operating parallel networks. Williams's VOP transport architecture will support voice-over-Internet protocol (VoIP) and voice-over-ATM (VOATM) by connecting to the local-exchange carrier (LEC) in a traditional fashion; the LEC networks will remain unaffected. The VOP transport architecture is illustrated in Figure 2. The voice stream is "packetized" using the VOP switch and routed over the optical transport layer via ATM and IP core switches, allowing for convergence of the voice and data traffic into one backbone. The broadband-services database is responsible for merging the telephony and data services into one server complex.
Williams's VOP transport architecture was designed to accomplish the following goals:
- Integration of voice and data traffic across the same infrastructure.
- Reduction of capital/operational costs.
- Provide a common set of services to the end users (whether data or voice customers), with integrated network-management and business systems.
- Reduction of point-of-presence (PoP) space requirements.
- Provide bulk bandwidth with a variety of protection schemes ranging from an unprotected wave (typically, a concatenated OC-48 rate connection that passes the k bytes and doesn't terminate the SONET/SDH line overhead) to 50-msec ring-like restoration for OC-3 (155-Mbit/sec), OC-12 (622-Mbit/sec), and OC-48 private-line circuits.
- Offer a scalable and efficient way of delivering narrowband services such as voice and non-SONET/SDH private line that does not hinder the development of the high-bandwidth services for data users.
From an architectural perspective, both IP and ATM are useful in a properly designed network. In local area networks, IP is nearly ubiquitous, evolving from the first implementation of the Ethernet standard. It is used in most corporate networks and campus environments, as well as on the Internet. The explosive growth in data networking has spawned the development of very-dense-capacity core routers (IP switches) capable of handling terabits of information. Thus, the proliferation of IP networks will continue to grow.
At the same time, however, the network industry has yet to agree upon and implement quality-of-service (QoS) functionality for an IP-based transport network that can be used in multivendor networking environments. Consequently, network designers are continuing to deploy an ATM transport layer between the SONET/SDH and IP layers in a vast majority of networks. The QoS controls embedded within ATM have proved to be very instrumental in providing congestion control and traffic management within carriers' private IP networks. QoS becomes more of an issue for VOP technology as VOP gains acceptance in the industry as a viable replacement for circuit-switched voice. Traditional telephony networks use equipment that provides network availability 99.999% of the time; therefore, interruptions of service seldom occur.
To ensure reliability and QoS for VOP, Williams uses its ATM backbone-transport network. To support the continual growth of IP traffic on the customer premises, VoIP at the access infrastructure of an IXC network is mandatory. Williams's VOP switching equipment handles this connectivity and connect to the customer using VoIP signaling protocols such as session initiation protocol (SIP) and H.323.
Continuing to support the circuit-based access infrastructure at the customer premises will also be a fundamental requirement during the migratory period for any IXC. To this end, the VOP switching equipment deployed by Williams will support customers accessing its IXC network using traditional TDM trunks. The trunks are controlled using either Integrated Services Digital Network (ISDN) primary rate interface (PRI) or multifrequency (MF) signaling.
The linchpin of providing telephony services at the IXC is the ability to interconnect with the public-switched telephone network (PSTN). Carriers such as Williams that choose to deploy VOP switching equipment must continue to support SS7 connectivity with the PSTN using traditional circuit-switched TDM trunking.
The service offering potential of a VOP network becomes a factor in the continual migration toward a seamless core network. The ability of IXCs to efficiently bring in network calling services for voice, such as toll-free, calling card, and debit card, via data-centric transport facilities further reduces the amount of network equipment needed in carriers' central offices. The broadband-services database-which supports interfaces such as Gigabit Ethernet and OC-3 ATM-is capable of delivering services to both data and voice customers. Data services may include but are not limited to dynamic IP and ATM virtual private network routing, ATM address translation services, IPv4 to IPv6 mapping, remote dial-up user verification, and connection to home office servers.
The business case for a VOP transport architecture is clear-bandwidth efficiencies, lower operational and capital costs, and reductions in network equipment, PoP space requirements, and support staff. Williams's VOP architecture will transport voice over the data network by packetizing the voice channels and statistically multiplexing the voice and data services within the virtual data cloud.
This approach will not compromise the service integrity that is standard today in circuit-switched, long-distance voice networks.
Mohammed Shanableh is a senior manager of network development, Kelly Houston is an engineer of network architecture, and Mark Allen is the director of optical networking at Williams Network, a division of Williams Communications (Tulsa, OK).