The Broadband Infrastructure

The Broadband Infrastructure

We are in an era of new, emerging networks that we loosely term next-generation networks. Data traffic in these networks is equal to or surpassing voice as the most mission-critical aspect of the network. Remember that when all the traffic is ones and zeros, everything is data, and voice is just another data application. Integration of voice, data, and video without protocol conflicts will greatly simplify the migration of legacy communication systems and network applications to next-generation transport technologies. The huge growth in e-business, extranets, and intranets will also require a convergent infrastructure that offers minimum latencies to ensure the responsiveness that customers need.

Traffic is growing at an alarming rate. More human users, more machine users, and more broadband access are all contributing to the additional traffic. Established carriers and new startups are deploying huge amounts of fiber-optic cable, introducing new possibilities, and optical technology is revolutionizing the network overall. This new era of abundant capacity stimulates development and growth of bandwidth-hungry applications and demands service qualities that can allow control of parameters such as delay, jitter, loss ratio, and throughput. Bandwidth-intensive applications are much more cost-effective when the network provides just-in-time bandwidth management options. Next-generation networks will provide competitive rates because of lower construction outlays and operating costs.

Converging Public Infrastructures

Public infrastructures are converging on a single set of objectives. The PSTN looks to support high-speed multimedia applications, and therefore it also looks to provide high levels of QoS and the ability to guarantee a granular diversification of QoS. The PSTN has traditionally relied on a connection-oriented networking mode as a means of guaranteeing QoS, initially via circuit switching, and now incorporating ATM as well.

The public Internet is also looking to support high-speed multimedia applications, and it must deal with providing QoS guarantees. But we are investigating slightly different options for how to implement this in the Internet than in the PSTN (see Chapter 5, "The PSTN"). Included in the IETF standards are Integrated Services (IntServ), Differentiated Services (DiffServ), and the new panacea, Multiprotocol Label Switching (MPLS), all of which are described later in this chapter.

Broadband Service Requirements

For next-generation networks to succeed, they must offer a unique set of features, including the following:

         High speed and capacity All next-generation networks must offer very high capacities, today measured in Tbps (terabits per second, or 1 trillion bps) and already moving into the range of Pbps (petabits per second, or 1,000Tbps). Higher-bandwidth broadband access (such as 100Gbps) will drive the need for additional core bandwidth, and discussions are beginning about networks providing exabits per second (that is, 1 billion Gbps).

         Bandwidth on demand Next-generation networks must be capable of providing or provisioning bandwidth on demand, as much as is needed, when it is needed unlike today's static subscription services.

         Bandwidth reservation Next-generation networks must be capable of offering reserved bandwidth, so that when you know you'll need a high-capacity service for streaming media, you can reserve the network resources so that they are guaranteed at the time and place that you need them.

         Support of isochronous traffic Isochronous traffic is timebounded information that must be transferred within a specific timeframe, and as such has a low tolerance for delay and loss.

         Agnostic platforms Agnostic devices support multiple data protocols (for example, IP, Frame Relay, ATM, MPLS) and support multiple traffic types, such as voice, data, and video, so that you can aggregate and administer all traffic at a single point.

         Support for unicasting and multicasting In unicasting, streams from a single origination point go directly to a single destination point. In multicasting, streams from a single origination point flow to multiple destination points. This reduces traffic redundancy by limiting the access to a selected group of users.

         QoS As discussed later in this chapter, next-generation networks must provide variable QoS parameters, must ensure that those service levels can be guaranteed, and must ensure that service-level agreements (SLAs) can be honored.

A number of developments have been key to enabling us to deliver on such a set of requirements. One such development is photonics and optical networking. Chapter 12, "Optical Networking," describes the revolution that started with the ability to manufacture glass wires, went further to introduce erbium-doped fiber amplifiers, grew to encompass wavelength division multiplexers and dense wavelength division multiplexers, and is proceeding forward to introduce optical add/drop multiplexers, optical cross-connect switches and routers, and the optical probes and network management devices that are very important for testing networks. We're looking forward to commencing deployment of the end-to-end optical environments in the next three to five years.

A number of broadband access technologies, both wireline and wireless, have been developed to facilitate next-generation networking. Chapter 13 covers these options, which include the twisted-pair xDSL family; hybrid fiber coax alternatives that make use of cable modems; fiber to the curb; fiber to the home; broadband wireless, including direct broadcast satellite, MMDS, LMDS, and Free Space Optics; and innovative new uses of powerline to support high-speed communications. As discussed in Chapter 15, "Wireless Communications," 3G wireless promises up to 2Mbps (but most likely around 384Kbps).

As discussed later in this chapter, multiservice core, edge, and access platforms are being developed. These platforms include integrated access devices (IADs), convergent switches, media gateways, agnostic platforms, and new generations of high-capacity terabits switch routers.

Intelligent networks, which include programmable networking, are potentially a replacement for the capital-intensive PSTN based on circuit switches. Softswitches, media gateways, Signaling System 7 (SS7) gateways, and service-enabling software developments are critical. (These developments are discussed later in this chapter.)

Characteristics of Next-Generation Networks

A next-generation network is a network that is designed for multimedia communications, which implies that it has broadband capacities, multichannel transport with high data rates, low latencies (80 milliseconds or less is the target), low packet loss (less than 5%, with the target being less than 1%), and QoS guarantees.

A next-generation network has a worldwide infrastructure that consists of fast packet-switching techniques, which make maximum use of transport and provide great transmission efficiencies. A next-generation network involves optical networking; today's electronic systems are going to be the bottlenecks to delivering tomorrow's applications, so we will see a replacement of the electronic infrastructure with optical elements that provide end-to-end optical networking.

A next-generation network has a multiservice core, coupled with an intelligent edge. The application of next-generation telephony in the edge environment may replace the existing architectures associated with the PSTN. Next-generation networks will be characterized by intelligent networking, for rapid service delivery and provisioning. They will also have video and multimedia elements, to deliver on the content for which the broadband infrastructure exists. Their access media are broadband in nature and encompass both wired and wireless facilities.

Next-generation networks stand to change how carriers provision applications and services and how customers access them. End-user service delivery from a single platform provides many benefits. It decreases time to market; it simplifies the process of moves, adds, and changes; and it provides a unique connection point for service provisioning and billing. Full-service internetworking between the legacy circuit-switched network and next-generation packet networks is mandatory going forward. Next-generation networks also must be interoperable with new structures that are emerging, which implies that they have to be able to support the most up-to-date transport and switching standards. They also must support advanced traffic management, including full configuration, provisioning, network monitoring, and fault management capabilities. In a next-generation network, it is important to be able to prioritize traffic and to provide dynamic bandwidth allocation for voice, data, and video services; this enables management of delay-tolerant traffic and prioritization of delay-sensitive traffic.

A next-generation network is a high-speed packet- or cell-based network that is capable of transporting and routing a multitude of services, including voice, data, video, and multimedia, and a common platform for applications and services that is accessible to the customer across the entire network, as well as outside the network. The main physical components of the next-generation network are fiber and wireless media, routers, switches, gateways, servers, and edge devices that reside at the customer premise.

Figure 10.1 shows an example of the kind of internetworking that needs to occur in providing access between the legacy PSTN and a next-generation packet-based network. It shows internetworking between the legacy PSTN and the underlying intelligent network, as well as emerging IP-based services. At the endpoints are customers that may be served by traditional plain old telephone service (POTS). They have access lines into their local exchange. The local exchange, then, has links into the SS7 infrastructure (discussed in Chapter 5), which allows service logic (for example, number verification, translation of toll-free numbers), the support of local number portability, and the provisioning of class services (for example, call waiting and call forwarding); these are the types of services and features that we're accustomed to with the PSTN. The local exchange connects into that intelligence to service the calls that are being requested by the users. The local exchange in a next-generation network also taps into the media gateway switch. The media gateway switch digitizes and packetizes the voice streams and then consults a softswitch, which is a form of intelligence in the network. The softswitch identifies the destination media gateway switch behind the telephone number being requested, and it applies to the packets an IP address that coincides with the destination gateway switch.

Figure 10.1. Internetworking between the legacy PSTN and a next-generation packet-based network

graphics/10fig01.gif

The gateway switches also connect into SS7 in order to be able to introduce or support the services that we traditionally associate with voice. Between the media gateway switches then, is the high-speed packet network, which could be an IP backbone, an ATM network, or an MPLS backbone environment. This network could also natively tap into new databases that are being created with advanced services designed specifically for the IP environment.

IP and ATM

IP and ATM are two techniques that are used in next-generation networks. (IP is introduced in Chapter 9, "The Internet: Infrastructure and Service Providers," and ATM is introduced in Chapter 7, "Wide Area Networking.") ATM is used to support a great deal of Internet backbones. Approximately 85% of all IP backbones use ATM in the core.

IP

IP was designed to work in the LAN world. It is a connectionless environment, which means it provides the capability of having information moved between network elements without a preconceived path between the source and destination. In a LAN environment, bandwidth is relatively inexpensive and the deployment, by definition, is over a small geographic area. Because of the small coverage area, transit delay is typically not an issue in a LAN.

In the event of congestion, IP discards packets. Transmission Control Protocol (TCP) retransmits the lost packets, quickly and transparently to the users, and because of the short transit delay, discarded packets are quickly detected, so users don't perceive delays in downloads. But WANs are typically deployed over longer distances than LANs. And in WANs, transit delays become a major issue in two ways: in controlling the QoS and in identifying the loss of packets that may have occurred because of congestion. Also, bandwidth is more expensive in a WAN than in a LAN; you pay for every bit that is sent over a WAN link, so packet discards that create retransmissions can make the expense of retransmission alone significant.

Problems with IP Networks Traditional IP routers were not intended to handle the large-scale type of networking that we are now demanding from IP. In IP router-based networks, the core, like the core in the PSTN, is responsible for providing interconnectivity. But in the IP router network, the core also provides server access and network management to the edge devices on the network periphery.

Because of the increases traffic networks are seeing today, the core network is becoming loaded, and that is resulting in network slowness and unacceptable delays. At the edge of the LAN, a shortage of network capacity, coupled with proliferation of broadcasts and multicasts, is creating significant network problems. When the edge demand exceeds the capacity of the core, queue overruns create capacity overload and lost packets, thereby reducing the availability and reliability of the network. As a result, users are suffering from congestion, inadequate server access, and slow response times.

Traditional IP routers cannot deliver the service quality that is increasingly being demanded. The shortcomings of traditional routers include poor path calculation and slow rerouting. Routers usually use the shortest path metric to calculate their routes, so IP routers send traffic over a shorter path, even if it's congested, instead of over a more desirable, longer, or uncongested path. This is one of the reasons that there is increased use of ATM or MPLS in the core for backbone traffic engineering purposes. Also, in the event of a backbone circuit or router failure, IP routers can take a long time up to a minute to calculate the new paths around the failure. This has led to more reliance on the resilient SDH/SONET backbone infrastructure, where there is a backup path a protect fiber that can ensure that the data is diverted to the protect fiber within a 50-millisecond timeframe.

Recent introductions of Voice over IP (VoIP) services and streaming media have exposed two other limitations of IP networks: latency and jitter. IP today doesn't provide a way to control latency and jitter. For a packet-based IP network to successfully support voice services, minimum transit delay must be achieved, as must minimum packet loss. High-quality voice demands less than 100 milliseconds for the total one-way latency, including all processing at both ends, which implies digitization, compression, decompression, queuing, playback, and so on, and that must also include the network delay. Voice compression and decompression alone normally take about 30 to 50 milliseconds. Network latency must be tightly controlled to support these services properly. One immediate solution is to increase the amount of available bandwidth. If there's no congestion, there's no problem. And technologies such as Dense Wavelength Division Multiplexing provide relief initially, but history has taught us that throwing bandwidth at a problem does not fix it, so this is a short-term relief measure, but not a long-term solution that addresses the need to differentiate traffic and its requirements on a very granular level. Hence the key to success for large-scale IP networking lies in delivering the flexibility of IP routing with a switched packet-forwarding mechanism that offers the highest possible performance and maximum control: IP switching.

IP Switching IP switching was designed to speed up increasingly choked networks, by replacing slower, more processing-intensive routers with switches. IP routers that provide connection-oriented services at the IP layer are referred to as IP switches. Routers are slower than switches because they must examine multiple packet fields, make substitutions in the packet headers, and compute routes on a packet-by-packet basis, which introduces latency and congestion.

The idea with IP switching is to make a connectionless data technology behave similarly to a circuit-switched network. An IP switch routes the first packet, and then it switches all subsequent packets. The goal is to make intranet and Internet access faster and to enable the deployment of new voice, video, and graphics applications and services. Therefore, IP switching has two objectives: to provide a way for internetworks to scale economically and to provide effective QoS support for IP. In essence, IP switching replaces Layer 3 hops with Layer 2 switching, which leads to good hardware-based forwarding performance.

Even with the advantages of IP switching, IP still doesn't allow us to properly administer all the QoS parameters that are part of traffic definitions, and this is where ATM comes in.

ATM

As discussed in Chapter 7, ATM was created in the WAN environment. It came out of the carrier community as a means by which to reengineer the PSTN for streaming applications. Because ATM comes from the carrier environment, where traffic engineering is essential, it is a connection-oriented technique. It provides a means to establish a predefined path between the source and the destination, which enables greater control of network resources. Overallocation of bandwidth becomes an engineered decision; it offers a deterministic way to respond to changes, on a dynamic basis, to network status. A great benefit of ATM is that it provides for real-time traffic management. It enables policing and traffic shaping; it can monitor (that is, police) the cells and determine, based on congestion, which cell should be dropped (that is, perform traffic shaping).

ATM allows networkwide resource allocation for class of service (CoS) and QoS provisioning. Again, because it is connection-oriented, it looks ahead to the destination point to ensure that each link along the way can deliver on the requested QoS. If it can't, the session is denied. Therefore, ATM also makes possible deterministic transit delay you can specify and calculate the end-to-end delays, as well as the variations in delays. This is all administered through multiple QoS levels (as described later in this chapter).

Remember from Chapter 7 that a lot of IP takes place over ATM. Because it is connection-oriented, ATM gives service providers the traffic engineering tools they need to manage both QoS and utilization. ATM's virtual circuits control bandwidth allocation on busy backbone routes. In provisioning a network, the service provider can assign each virtual circuit a specific amount of bandwidth and a set of QoS parameters. The provider can then dictate what path each virtual circuit takes. Basing these decisions on overall traffic trends reduces the likelihood of network hot spots and wasted bandwidth, and this is why so many service providers turn to ATM to transport IP traffic. However, the service provider has to deal with two control planes managing both IP routers and ATM switches. Using ATM virtual circuits to interconnect IP routers leads to scaling problems because every router needs a separate virtual circuit to every other router. As the network grows, the number of routes and virtual circuits can increase exponentially, eventually exceeding the capacity of both switches and routers. Network operators can work around this in one of two ways: either they can forgo a full-mesh architecture or they can move to MPLS, which is discussed later in this chapter.

Table 10.2. IP Versus ATM

Transport

Benefit

Drawback

Services Supported

Packet size

Header

IP

Pervasive at the desktop

No QoS

Data, voice

Variable, 40 bytes to 64,000 bytes

40 bytes

ATM

Multiple service classes

Small cell size that is inefficient for data transport

Data, voice, IP, Frame Relay, X.25, leased lines

Fixed cells, 53 bytes

5 bytes

IP Versus ATM

Table 10.2 is an overview comparison of IP and ATM. An upside of IP is that it is pervasive at the desktop. The downside is that there is no QoS built in. It supports data, voice, and fax. IP packet size is variable. It can be up to 64,000 bytes, but packets are segmented into 1,500-byte frames for transport, and 40 bytes of each packet is for the header information.

The upside of ATM is that it is an architected QoS approach that defines five key service classes (described later in this chapter). The downside is that it uses a small cell size (only 53 bytes, 5 bytes of which is the header information), which means it has a lot of overhead (that is, cell tax), which could be construed as inefficient for data transport, or for voice transport, or for other traffic types and this is an issue when bandwidth is constrained and expensive. Remember that ATM was built based on the assumption of gigabits-per-second trunks and generous bandwidth, so the cell tax is less relevant if the prevailing condition is the unleashing of abundant bandwidth. ATM supports a wide variety of services, including voice, IP, Frame Relay, X.25, and leased lines.

Terabit Switch Routers

Terabit switch routers are relatively new, and they are an emerging class of backbone platform. They will offer, as the name implies, terabits of capacity the current products range from some 640Gbps to 19.2Tbps and they will support interfaces that range from OC-3 (that is, 155Mbps) to OC-192 (that is, 10Gbps).

Terabit switch routers are agnostic devices, so they support a wide variety of data traffic types, protocols, and interfaces. They are engineered for short and predictable delays, and they offer robust QoS features as well as multicast support and availability in carrier class, so they will be able to service a wide subscriber base.

Terabit switch routers are being developed to integrate with other network elements, in particular optical switches, and they may communicate by using MPLS. Again, these are the early days for terabit switch routers at this point the platforms are largely proprietary and vary in how they operate.

We don't really have to choose between IP and ATM. At least for the time being, we can use them together quite effectively. IP has become the universal language of computer networking, especially in the desktop environment. IP-based services including VPNs, e-commerce, outsourced remote access, application hosting, multicasting, and VoIP, along with fax and video over IP are used in a number of areas. A benefit of IP is that there is a much larger pool of knowledgeable programmers for IP than there is for ATM. However, all these wonderful applications that the programmers are developing for IP tend to require a lot of CoS and QoS, as well as controlled access. IP standards for QoS are in the early development stages. ATM will increasingly be used to switch IP traffic because of its network management, restoration, and reliability capabilities.

 



Telecommunications Essentials
Telecommunications Essentials: The Complete Global Source for Communications Fundamentals, Data Networking and the Internet, and Next-Generation Networks
ISBN: 0201760320
EAN: 2147483647
Year: 2005
Pages: 84

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net