Mobile operators are rapidly deploying 4G technologies such as WiMAX in an effort to increase the mobility of high-speed Internet access. This emerging deployment of high-capacity broadband wireless will enable a new set of customer services focused on the convergence of data, video and voice. All of this activity is driving a new set of backhaul requirements to enable these services.

The emerging services are driving a few fundamental network core requirements: high capacity, data-centric, ability to offer voice, and data. This drives a number of requirements on the backhaul architecture. The data orientation and Ethernet standards that these systems are based on is a major departure from the existing T1 backhaul that exists for DSLAM or cellular tower locations, driving a data-centric backhaul network. Additionally, the bandwidth-intensive applications that are enabled through WiMAX will drive much higher bandwidth than the 5-10 Mb/s that had previously been required. In addition, because this is a new architecture with new applications, the bandwidth is expected to grow quickly, and drive new revenue, which requires a backhaul architecture that can rapidly scale. From a performance aspect, the ability to offer multiple services such as voice and video requires that the backhaul architecture provide low latency and jitter. Lastly, the services being offered must fit into the existing price structure of data services, and therefore it is fundamental that the entire architecture is low CAPEX, with minimal recurring costs. The resulting set of requirements on the backhaul architecture is:

• High capacity: To enable broadband mobile service
• Rapid scalability: Required to handle rapid market capture
• 99.999% Availability with 50 ms switching: For real-time critical services
• Ultra-low latency and jitter: Required for voice and video service
• Native Ethernet: Required for WiMAX data architecture
• OPEX and CAPEX Efficient: To enable profitable 4G business

To meet all of these requirements, legacy T1 architectures must be abandoned, as they do not meet the data needs, OPEX or scalability requirements of 4G services due to the following limitations:

• Recurring leasing costs
• Capacity and scalability limitation (E1/T1)
• TDM, not Ethernet based
• Line availability (or time to deploy if unavailable)
• ROW issues
• Building entrance consideration
• Potential construction delays

To address the backhaul requirements of 4G services, a new backhaul architecture must be adopted. Ideally this architecture could be all fiber, however the cost and rapid deployment timeframes limit the feasibility. Instead the existing fiber infrastructure must be leveraged and extended wirelessly. To address this, there is an emerging architecture consisting of wireless backhaul. There are a few key characteristics to this emerging architecture:

• Licensed: Both in the common carrier (11,18,23 GHz), and the area licenses (24, 28, 38 GHz). This provides the insurance of network availability, and predictability of service. In addition, this spectrum enables capacities beyond 500 Mb/s for maximum service scale.
• Native Ethernet: To enable emerging data services
• Mesh/ring architectures: To cost effectively deliver 99.999% availability with 50 ms switching
• All-outdoors: To minimize indoor OPEX lease costs

The emerging architecture consists of two layers as shown in Figure 1 below:

• Ring/mesh core
• High capacity core links (100M-800M)
• Ring/mesh topology for high availability
• Common carrier licensed links for long reach
• Hub and spoke access layer
• Lower capacity access links (<100M)
• Shorter unprotected links
• Area license or common carrier licenses

Figure 1: Example Wireless Backhaul Architecture

In the following sections, we will examine in detail, each of the key parameters of this architecture:

Availability: The two contributing factors to wireless unavailability are the equipment and the path unavailability. The equipment unavailability can be virtually eliminated by using a ring/mesh architecture, which will always provide a redundant equipment path, bring the effective equipment availability to >99.999999% (negligible unavailability). In order to increase the service availability to 99.999%, path is required. For wireless links, the major factor effecting availability is the rain, however this can be reduced by providing diverse paths. The path diversity improvement factor (PDIF) is a measure of the joint probability of two co-joined links failing simultaneously. A PDIF of 5 or higher is quite common for a link of 10 Km. This results in the unavailability of the link being reduced by that factor. The service availability of a 99.99% path becomes 99.995%, and the unavailability of a 99.995% path becomes 99.999%. These two factors combined with licensed links, which eliminate interference, enable the core to cost effectively be engineered to 99.999% availability. Ring, 1+1 or unprotected links can then be selected on a per site basis for the access sites, depending on the individual availability requirements.

Capacity and Scalability: The capacity of the core ring can scale up to 800 Mb/s with many licensed systems. Assuming 100% protection bandwidth is provisioned, this provides 400 Mb/s of working and 400 Mbps of protection in each direction around the ring, enabling 800 Mb/s of working traffic in total. In the example architecture in Figure 1, there are 23 end sites, so this architecture will provide 35 Mb/s of scale per site. Assuming 50% oversubscription of the protection bandwidth (services would be reduced in throughput during failure), 50 Mb/s per site is supported. If scale beyond this is required, the rings can be overlaid, or can be split, to double the capacity. Assuming that line of sight is available, the ring splitting option is preferred, as it is the lowest cost option, by only requiring a single link overlay. These two methods are shown in Figure 2 below.

Figure 2: Ring Capacity Doubling

In the access layer, links can typically scale via software from an initial capacity of 10-20 Mbps up to 100 Mb/s or higher.

Latency and Jitter
The number of intermediate hops and switches determines the overall end-to-end latency. The average latency of each hop is 150 microseconds and the average switch latency of an Ethernet switch is 100 microseconds. In Figure 2, the example highlights the shortest path of two links, this yields a latency of .4 ms. For the longest path of 5 links, there would be an average latency of 1.5 ms. In addition with the wireless network, the latency is very predictable due to the licensed operation, and very predictable latency of licensed links. This is well below the 4G latency budget, and far below leased lines which can have wide variances from 1m to > 10 ms, depending on length and equipment being traversed.

OPEX and CAPEX Efficient
Beyond the performance characteristics above, it is fundamental that the architecture provides cost effective backhaul, so the 4G business case will work. In order to examine the cost effectiveness, we will compare the cost of building a wireless network against the cost of leasing Ethernet services (assuming that Ethernet services could be leased). For this comparison, we will look at the model shown in Figure 1, with 5 ring nodes and 23 access sites. This comparison includes CAPEX, link installation, licensing costs, monthly antenna lease costs, Ethernet switch costs, and an annual maintenance charge as shown in the table below. The costs of building a wireless network are compared against a monthly 10M lease cost of $1000.

Initially, the comparison will be for a 20M service per site. In this comparison, there is a payback period of 13 months as shown in the graph on the left in Figure 3. We then perform a sensitivity vs. capacity comparison as shown at the bottom of Figure 3 below, you’ll see that the payback period goes down to 6 months as the capacity per site increases.

Figure 3: Leased line vs. Wireless Network Cost

Wireless backhaul architectures offer operational improvements over leased lines that are harder to quantify, but still very important. This includes the ability to manage your own service levels, and the delivery and scale of circuits. In addition, by deploying an all-outdoor architecture, indoor infrastructure costs and requirements can be minimized.

Conclusion

The rapid deployment of 4G services and their new set of requirements and scalability are driving the need to evolve mobile backhaul away from traditional leased T1s. The wireless backhaul architecture can be rapidly deployed to quickly deliver 4G services, and provide the capability for rapid scalability from 10M to greater than 100M per site. Using next generation Ethernet services, the wireless architecture provides backhaul with <2 ms latency. At the same time, service availabilities can be engineered from 99.95% to 99.999% with carrier-grade 50 ms switching. Most importantly, this architecture is extremely cost efficient delivering payback period of <1 year vs. leased Ethernet services.

Greg Friesen is Director of Product Management for DragonWave Inc.

Visit DragonWave Inc. online.