IP's Rough Edges

It's no surprise that explosive growth in Internet use will fuel an increase in telecommunications traffic. Although Internet and other data are poised to make up the bulk of traffic within the next few years, the lion's share of revenues still will come from traditional services such as voice.

For broadband carriers, such as those holding LMDS spectrum in the 28GHz band, these factors play a key role in their architecture choice. These carriers must address traditional revenue-generating services while deploying an architecture that supports high-growth future services.

Given the undeniable increase in Internet use, IP is emerging as the dominant protocol in both public and private networks. It also is becoming the convergence protocol for all traffic types, including traditional services, such as voice telephony. IP, with its suite of protocols, can efficiently transport various data types over the same network resources; however, IP traffic is bursty, best-effort in nature and needs to be routed hop by hop. These shortcomings present challenges in deploying IP networks and in providing business customers with a predictable quality of service (QoS). Many standards groups are adding quality and class-of-service capabilities to the IP protocol, but it may be awhile before they are available and widely deployed.

Using ATM technology as a transport mechanism has matured and gained acceptance in the broadband wireless market as another way to deliver multiservices over the air (OTA). ATM was designed to switch traffic quickly and efficiently and to offer predictable QoS.

IP can exploit ATM's switching-transport and traffic-management attributes within the broadband wireless network to deliver predictable end-to-end QoS by adapting IP to ATM at the network edges. ATM traffic shapers smooth out the bursts of IP data, while the ATM connections, configured with specific peak, sustained and minimum-rate traffic parameters, switch the traffic through the network at guaranteed QoS levels to conform to contracted service level agreements (SLAs).

ATM-BASED LMDS The enormous amount of bandwidth inherent in LMDS enables carriers to offer a broad range of services and opens up new markets for remote LAN connectivity and routed IP interface services. The distinction between WAN/LAN interface is blurred, since the WAN now is capable of affordable LAN speeds.

A typical metropolitan network topology has LMDS base stations structured around a core base station that also serves as a network hub. The base stations in the network enable the customer sites within each cell to connect to the various services within the network or to other parts of a larger WAN.

Each base station can service many customers' provisioning connections that fall into the following general service categories:

* Fixed bandwidth (T1s/E1s or fractionals, OC-3)

* Switched bandwidth (switched and concentrated voice circuits)

* Dynamically varying bandwidth (10/100baseT, FDDI).

To handle multiple types of services effectively, the network configuration includes:

1. An ATM backbone to provision the various bandwidth requirements and QoS attributes of the multiple types of services. ATM allows the network operators to provide service-level guarantees to their customers for each type of service.

2. TDMA wireless links with dynamic bandwidth allocation (DBA) that are capable of delivering both fixed and dynamically varying bandwidth connections while allowing for traffic multiplexing from many customers at different locations at a fine granularity over the same RF carrier. This setup lowers the cost of providing network access to customers with smaller bandwidth requirements by sharing the downstream and upstream RF carrier among multiple CPE sites and allows for bandwidth overbooking for better use of network resources.

3. Frequency division multiple access (FDMA) for high-speed links between network points that require large amounts of bandwidth, such as for inter-cell links or customer sites that require the full bandwidth of an aggregate link. A wireless intercell link allows the provider to bypass the local fiber wireline carrier to connect its base stations.

IP PROTOCOL IP is the most widely used protocol in both business and public LANs and WANs. It also is becoming the de facto convergence standard for all traffic types, including traditional and future telecommunications services.

The IP protocol in its base form has no mechanism to ensure QoS, due to its connectionless nature and best-efforts routing process. Standards such as Diff-Serv and MPLS (multi-protocol label switching) are evolving IP with QoS and connection-oriented capabilities. The Diff-Serv Working Group is redefining the "type of service" field found in the IP packet header to provide a way to offer differentiated classes of service, support various types of applications and meet specific business requirements. The MPLS Working Group aims to add a packet header to the IP packet that specifies how to route the packet through the network.

Today's approach to the IP networks' QoS problem is to provision more bandwidth to handle larger bursts of data and to perform IP-routing operation in hardware to reduce congestion in the core routers. This system has worked in most cases, but with the increase in public Internet use, the high processing power of PCs, a growing demand for service-level guarantees and more killer applications requiring high bandwidth, these solutions are not future-proof.

When considering a large routed IP network in which many routing nodes are traversed to forward traffic to the destination, each packet sent must be examined at each hop in the network. This complex process at each node introduces delay and delay uncertainty because the processing involved employs packet queues of varying lengths. Specific network implementations can prioritize the queue ordering to improve certain aspects of performance, but true performance guarantees are impossible to enforce, making SLA administration also impossible.

This process aggravates general congestion and queuing problems (delay and delay variability), which makes QoS administration difficult. The performance attributes of time-sensitive traffic streams, such as voice and video conferencing, are of particular concern.

OVERBOOKED & BURDENED An important aspect of packet- or cell-based networks, such as IP and ATM, is overbooking network resources to improve bandwidth usage and increase revenues for the carrier. The bursty nature of IP traffic allows statistical gain to be achieved with a number of users and applications sharing network resources. The ability to share bandwidth while still providing customers with a high-speed access interface for bursty traffic is the key to lower operational costs and efficient network resource use.

However, overbooking resources may lead to packet loss when traffic sources produce sustained bursts of data that exceed the network resource capacity. The solution is to provide traffic shaping at the traffic's source to conform to a contracted SLA. Doing so effectively paces out the bursts of data at predetermined rates and allows network operators to engineer the network to meet service-level guarantees and use network resources at optimal efficiency.

When considering IP over a broadband wireless network, customer sites are connected to the network over fixed-bandwidth leased lines, such as T1/E1, with appropriate interfaces at the customer site. Traffic from customer sites is aggregated at the base station, then sent over high-speed links to a central switch connected to a core IP router, which routes the traffic between cells, the Internet and other portions of a larger network. This architecture presents several challenges to the carrier:

* Poor scaleability: The network backhauls all the IP traffic to a centralized node where core routing takes place. As the network expands and more customers are connected, the processing capacity and the number of interfaces that the core router can handle becomes a bottleneck and limits the network's scaleability.

* Unpredictable performance: The core router's performance depends on the network traffic for all users, which makes it impossible to offer service guarantees.

* Inefficient traffic flows: Back-hauling all the traffic can congest the backhaul links, which may require more backhaul links or larger links to remedy.

* Inefficient use of core and air bandwidth: Using simple base-station structures, such as FDMA links to the customer premise and leased line services to the customer, does not allow significant statistical gain OTA and in the backhaul links.

* Poor overbooking of network resources: Base stations without integrated traffic-shaping functionality cannot effectively exploit the statistical multiplexing advantages of links carrying IP traffic. Even large corporate IP connections generally exhibit bursty attributes that can be exploited to yield significant capacity benefits OTA and back-haul links. In this particular case, poor efficiencies may result.

IP OVER ATM-BASED LMDS An LMDS network that combines IP with ATM can provide switched IP routing and classification of IP packets to ATM virtual circuits (VC). If it integrates TDMA/DBA with ATM, it can to maximize the OTA bandwidth.

In the switched-routed network, the complex processing operations required to route the packet are performed once, at the network's edge, to select an ATM connection to the destination. Subsequent nodes switch packets between interface ports. The ATM connection may be a permanent virtual circuit (PVC) or a switched virtual circuit (SVC). PVCs are fixed connections provisioned by the carrier. SVCs are created dynamically on demand. SVCs are preferred for network scaleability and minimum operator configuration and maintenance.

In order to support QoS with IP traffic, IP packets need to be classified and mapped onto the ATM QoS framework at the IP/ATM service interfaces at the network's edge. In addition to providing the rules for handling various IP prioritization, this map forms the basis for the SLAs offered to customers purchasing IP services from the LMDS carrier. To capitalize on the ATM-transport-layer performance, this function is pushed as far to the edge as possible. In current LMDS network implementations, it is possible to push this functionality into the customer premises equipment, which also allows highly efficient MAC implementations to be exploited OTA.

When LMDS network topology is integrated into the ATM network, IP traffic is classified and prioritized at the customer premise. The traffic is sent to the local base station over VCs that meet the QoS requirements of the IP packets. The VCs are terminated on the base station, where routing information establishes SVCs tothe destination base station or external network. SVCs then are used to switch the IP packets through the network at the required QoS level.

This network model has the following advantages:

* Scaleability: Complex routing is performed at the network edge, and SVCs are used in the network core. There is no core router as the bottleneck. The routing operations are distributed throughout the network.

* Multiservice customer offering: ATM is capable of transporting various data types, such as IP and TDMA traffic, with equal ease and efficiency.

* Guaranteed QoS: Able to administer and police QoS attributes on large numbers of connections in order to support SLAs.

* Efficient traffic flows: Switched routing allows IP traffic to take a direct route to the destination instead of funneling to the network hub.

* Efficient use of OTA bandwidth: ATM switching fabric integrates with TDMA OTA to provide high-efficiency bandwidth use and the ability to provide DBA and switched-circuit operation.

* Overbooking of network resources: ATM traffic shapers can smooth out the IP traffic bursts to conform to the SLA and minimize packet loss while allowing maximum reuse of network resources.

* Minimal network operator intervention: Using SVCs reduces the management of setting up connections and the connection maintenance.

LMDS and other broadband carriers are able to exploit their greenfield advantage by choosing an architecture that enables them to maximize current revenue opportunities while supporting advanced and future services. To address emerging IP services market opportunities, the carrier needs a network that can provide connections with QoS guarantees, allows for resources overbooking, offers a broad range of services and interfaces, and can scale to support a large number of customers. An LMDS network topology that can provide these capabilities features IP and TDMA/DBA integrated with an ATM transport. This network can offer IP services that cannot be implemented on traditional router-based networks. The network also can scale in size and bandwidth to meet the requirements of growing demand for IP applications such as voice telephony, video conferencing and video multicasting over IP.