Integrated Access - Laying the Groundwork for the Networks of the Future

The move to converged access is having a fundamental impact on the broadband industry for both operators, who want to provide integrated voice, data, video, and entertainment over their networks, and vendors, who are developing products to support this new infrastructure. An examination of how each of the major broadband technologies (DSL, cable, wireless, satellite, and fiber) supports integrated access services is neccessary for a glimpse into the future public network.

Advances in DWDM (dense wavelength division multiplexing) technology are rapidly increasing the bandwidth that can be transported over fiber optic cable. Meanwhile, the explosive growth of Internet and IP (Internet Protocol) based traffic — including bandwidth hungry multimedia content — is driving demand for broadband transport in access networks. To capitalize on this market opportunity, network operators are looking to deliver multimedia over IP, as well as many other kinds of interactive entertainment and informational services.

As the relative portion of IP traffic increases, the next logical step for carriers is to optimize the public network for IP transport. By the end of 2004 more than 95 percent of the traffic volume (number of bits) transported over public networks is expected to be generated by applications that are running on top of the IP protocol. Today, the majority of new applications are being developed for IP-based transport, and the desktop is totally ruled by IP. It is generally accepted that all services (voice, video, data, and multi-media) are converging towards a single IP-based infrastructure.

Evolution Of The PSTN

The public switched telephone network (PSTN) can be divided into two layers: The core network and the access network. The fact that IP will vastly dominate traffic volumes implies that the most cost efficient network architecture is one that is optimized for IP. Therefore, it is expected that the core network will migrate towards a two layer protocol stack containing an optical layer and an IP/MPLS (multi-protocol label switching) layer. This protocol stack is predicted to represent the majority of new public multi-service network deployments by the year 2004.

The two-layer, IP over glass, protocol stack is in use today for producing best effort type Internet services. A true multi-service network requires advanced QoS (Quality of Service), traffic engineering and traffic protection capabilities that are not present in most legacy IP routers.

Today QoS and traffic engineering are provided by an ATM (asynchronous transfer mode) layer and traffic protection is provided by an SDH/SONET (synchronous digital hierarchy/synchronous optical network) layer. The optimization of the network means that the IP and optical layers will need to absorb these functions as the need for QoS, traffic engineering and protection will only intensify as new latency sensitive applications, including voice, are added.

As soon as QoS, traffic engineering and multi-service transport are supported by the IP/MPLS layer, network operators can eliminate the ATM layer, as it is no longer needed. Note that eliminating the ATM layer only means getting rid of the ATM control plane. The physical ATM devices and cell-based transport systems can be integrated into the IP/MPLS network by implementing an MPLS control plane in the existing ATM switches.

SONET/SDH will remain as an underlying layer since the inclusion of protection capabilities in the IP/MPLS layer is a difficult task. These legacy TDM (time division multiplexing) networking technologies can execute protection switching in time scales that are measured in tens of milli-seconds. Despite claims to the contrary by some IP/MPLS equipment suppliers, this is not possible today with IP/MPLS layer protection mechanisms.

The solution to the protection problem will be two-fold. First, it should be recognized that most data applications do not require rerouting to take place in tens of milli-seconds. This will allow mechanisms that are based on network layer signaling to be used for protection of data traffic. It is expected that these types of mechanisms can achieve rerouting within a couple of seconds of downtime in realistic networking failure scenarios. While a couple of seconds is sufficient for most data applications, it is not acceptable for voice or other demanding applications. Therefore, specific protection mechanisms will be needed that are capable of transmitting traffic over two separate paths. These types of protection mechanisms will be required to allow the full migration to a pure IP over glass protocol stack."

The Physical Layer of the Access Network

What is the most future proof technology for the physical layer in the access network? For large enterprises the best access solution is point-to-point fiber. DS1 leased lines will remain an important access media for medium-sized businesses, however cost prevents their use for smaller enterprises and residential customers. A combination of several alternative technologies, including DSL (digital subscriber line), HFC (hybrid fiber coax), wireless radio, optical fiber, and satellite transmission, is more suitable for these markets.

ADSL: Copper is the Weak Link

ADSL (asymmetric digital subscriber line), which utilizes existing telephone cables for high speed data transmission, can provide up to 8 Mbps of bandwidth downstream and up to 1 Mbps upstream. While these numbers sound great, the bandwidth that service providers actually guarantee for current DSL offerings is only a couple of hundreds of Kbps or at most 1.5 Mbps. The reality with DSL is that the quality of current copper access networks is not good enough to enable operators to support bandwidth levels necessary for real multi-media services.

Running good quality digital video over IP transport requires 3-10 Mbps of bandwidth and there is a desire to support multiple simultaneous channels. Guaranteeing bit rates at this level will require operators to deploy fiber closer to the customer and will use copper plant only over the last one to three thousand feet. Once the active DSL infrastructure is brought this close to the customer, ADSL is no longer the most attractive DSL technology. Instead, operators will most likely deploy VDSL (very high bit-rate digital subscriber line), which provides up to 50 Mbps (the real service offerings will most probably be in the 10-25 Mbps range) of bandwidth downstream and up to 13 Mbps in the upstream direction.

The legacy public switched telephone network (PSTN)HFC: Lacks Scalability

Currently HFC, which is used for distributing CATV (cable television) signals, is the most widely deployed broadband access technology. For broadband IP, the HFC infrastructure typically provides 10-35 Mbps of downstream bandwidth and 2-10 Mbps of upstream bandwidth. Unfortunately, HFC is a shared infrastructure with a fixed pool of bandwidth typically accessed by hundreds of subscribers. This provides very good support for current web applications, but as the IP infrastructure is increasingly used for the delivery of multimedia content, it is questionable whether HFC can really scale well enough to satisfy growing bandwidth demands alone.

Wireless/Satellite

Wireless technologies are attractive because they are quick to deploy. Unfortunately, high equipment costs and line-of-sight requirements are significant impediments to wireless broadband deployments. In the end, the cost of wireless solutions for providing large-scale broadband IP access with multi-megabit speeds is very high. Satellite-based solutions are attractive for broadcast type applications. However, inherent delay and cost considerations limit their use for interactive communications.

PON: A Long-Term Solution

Fiber-based access networks provide almost unlimited bandwidth, are not susceptible to electro-magnetic interference and provide very consistent service to all subscribers regardless of local-loop length. Traditional fiber technologies (SONET rings and DWDM) have been designed for the needs of backbone networks. Fiber-based residential (and small and medium-sized business) access deployments will likely be based on PON (passive optical network) type deployments. PON uses passive optical splitters and combiners for sharing the fiber distribution network among multiple users and one PON is typically shared between 32-64 customers. Today, the downstream bit-rates are 155-622 Mbps with similar bit-rates in the upstream direction.

A big advantage of PON is that once the fiber is installed bit-rates can be scaled to multiple gigabits per second and even beyond that. There is no need to replace the distribution plant with new technology in the foreseeable future. The disadvantage of PON is the cost of laying fiber. If a network operator already owns copper infrastructure, VDSL usually proves more economical.

VDSL and PON on the Physical Layer

Clearly VDSL and PON are the two best alternatives for delivering broadband IP access to residential and small business customers. Although the large scale deployment of these technologies requires significant new investments — PON requires operators to lay new fiber all the way to the subscriber and VDSL requires the installation of new DSL multiplexers closer to customers — HFC and wireless technologies lack the scalability to support the delivery of multi-media IP traffic. The first PON and VDSL deployments are taking place now, with explosive growth expected for both of these access technologies in a couple of years time.

Earlier we discussed the evolution of the core of the new public network into two layers: an IP/MPLS layer and an optical layer. It is also predicted that the control plane for the optical layer will be based on MPLS protocols. If the control plane for both layers of the core network is IP/MPLS, it follows that the optimal control plane for the access portion of the network should be IP/MPLS based as well.

One homogenous control plane end-to-end is the most logical and efficient way to build and manage a network. Using the same set of tools throughout the network minimizes the headaches of interoperating different pieces of equipment.

Today's access network is primarily based on TDM. However, most people agree that the rigid structure of TDM is not optimal for serving IP-based applications whose bandwidth demands are dynamic and constantly changing. The use of TDM to serve these types of services implies that bandwidth for each service has to be provisioned according to the peak requirement of the service and bandwidth cannot be shared between services. On the other hand, packet-based integrated access enables dynamic bandwidth allocation and statistical multiplexing, which reduces costs and renders the TDM-based alternative uneconomical.

The only real competitor for an IP/MPLS-based access network is an ATM-based alternative. The main disadvantage of ATM is its lack of integration with the IP control plane. With ATM, a service provider must maintain and manage two separate control planes, which complicates the network. ATM is also a very inefficient transport mechanism for IP and wastes approximately 10 percent of available bandwidth compared to an MPLS-based approach. Furthermore, because ATM uses a totally different QoS model than IP, it prevents network operators from deploying seamless, end-to-end QoS schemes between access networks and the IP-based core.

IP Data Services Residential Services

For the foreseeable future, the most popular data service is likely to remain Internet access. This service will one-day be ubiquitous in much the same way as current voice service offerings. Currently, the Internet bandwidth that can be reliably delivered to an individual user is measured in tens-of-kilobits per second (this was not a typo, it is surprisingly low). The real explosion in residential multimedia services is going to take place once service providers can reliably deliver high-quality video over IP. Going from the current tens of Kbps to the required Mbps (a 100 fold increase) of bandwidth is achieved by deploying true broadband access technologies and bringing content closer to the user.

Business Services

For businesses, Internet access will evolve from the current best effort service to a service portfolio that offers multiple traffic classes. While Internet access will be the highest traffic generator in the new public network, the largest data revenues will be generated by IP virtual private network (VPN) services. These services will replace the current leased line and Frame Relay networks.

Once QoS aware IP VPNs begin getting deployed, the distinction between a voice service and a data service will be eliminated. The same IP fabric is used for switching both voice and data.

The new public network must be able to support numerous access topologies over a unified protocol.

The combination of IP-based applications running over a ubiquitous IP infrastructure will make it easy and economical to outsource application hosting and IT functions. ERP (enterprise resource planning) application hosting, web-based data storage and other ASP (application service provider) offerings are big growth markets that stand to benefit from IP-based access. Even voice service may become just another IP application in the soft-switch model, which could be delivered by a voice ASP.

Voice Service

Despite all the hype that surrounds the Internet and IP, IP services are not yet significant contributors to the overall revenues of telecom service providers on a global scale. In 1999, legacy voice services generated more than 90 percent of service provider revenues worldwide. IP services (Internet access, IP VPNs, mobile IP, etc.) were still in the low single percents (estimate 2-4%).

One of the main benefits of integrated access is its ability to support revenue generating legacy voice services over the same infrastructure that provides the multimedia services of the future. The current PSTN incorporates a huge amount of voice service functionality and represents a large capital investment which cannot be replaced by IP solutions over any short time period. Forecasts indicate that the circuit switched legacy voice network will continue to grow over the next five years. In legacy voice networks, the role of IP is currently limited to providing transport between TDM switching elements, with all services delivered, and all switching executed, using TDM equipment. In the short term, the role of integrated access will be to bring traffic over a packet-based access infrastructure to the legacy TDM switches.

Even though the existing PSTN will dominate voice service delivery over the next five years, IP-based voice services will grow steadily over the same period. IP voice is expected to start generating significant revenues by the end of the five-year period. Some forward looking operators will deploy IP-based voice switching solutions during the first half of the five-year period.

Next Generation Public Network

In the near term, the legacy PSTN and new world IP/MPLS-based public network will co-exist side-by-side. Eventually, the deployment of MPLS-based access networks will lead to an end-to-end network with one homogeneous IP/MPLS-based control plane that spans from the customer premise across an optical DWDM core, and on to the end user through a variety of broadband access methodologies. This new architecture will spell big savings for network operators since all the network layers can be managed using the same family of protocols. It will also provide an integrated platform for creating new revenue generating data and unified messaging services that can be provisioned and activated with speed and flexibility.

Top of Page