The Economics & Practicality of Terragraph:
Is Terragraph Backhaul essential for 5G?

Executive summary

On April 25th, 2018 AT&T and the city of San Jose have reached an agreement to install a network of 170 small cells on lampposts with plans to increase to 1,000 sites over the next 2-3 years. This network deployment will be one of the earliest 5G mmWave access networks in the country. As this network expands one of the biggest issues AT&T will likely face is access to fiber at each lamppost. In this case study, we examine the practicality of using Terragraph to meet a tier 1 operator, like AT&T, needs of a rapidly growing or high-density deployment. Terragraph is a line of sight (LOS) based mmWave backhaul that operates at 60GHz.

This case study will cover the following topics:

• Feasibility of Line of Sight between street lights in San Jose

• Terragraph backhaul performance

• Operator’s 5-year rollout plan

• Economic analysis covering the capex and opex cost

In order for Terragraph to be a viable solution, it must first have line of site. The results of this study find that in a city like San Jose it is feasible to get line of sight between streetlights with an average distance of 73m. This would allow operators to easily support capacity requirements at 5G sites with fewer fiber points of presence (PoPs).

The greatest cost driver of 5G is the backhaul. The economics analysis portion of this study demonstrates that an operator can realize significant cost saving in backhaul by deploying Terragraph. As shown in Figure 1 the Terragraph backhaul Network Expense breaks even in 2020 with increasing savings as the network grows.

Los FEASIBILITY study

In order to establish line of sight we use location data for 63,00 street lights and LiDAR data to map the presence of obstructions including buildings and trees for Santa Clara County. The pictures below clearly show the line of sight combinations possible along with obstruction due to buildings and trees. In the downtown picture, there is an entire street with no line of sight possible due to the presence of trees.

The data set representing the distance between street lights with LOS is shown below as a histogram. On average, the distance between street lights with LOS is 73.3m with small percentile approaching 250m LOS. At these distances of 250m and less, there will be no issue in maintaining a 60GHz mmWave link. San Jose with clear LOS between large number of street lights is a good market for Terragraph deployment.

Terragraph Network Performance

AT&T’s initial rollout has been announced to be 170 sites, but the memorandum released by the city of San Jose shows 181 sites spread across 10 council districts. For modeling purpose, this study assumes 181 sites picked randomly from the existing street light locations while remaining consistent with the council district specific plans. These 181 locations represent the starting point of AT&T’s network which will be fully deployed by end of 2018. Since this is a greenfield deployment by AT&T, it’s fair to assume that each of these 181 sites will have access to fiber along with a mmWave Small Cell.

End of year 2018, with 181 sites represents AT&T’s baseline network. Starting year 2019 an operator like AT&T will have two choices, either use Terragraph for their backhaul or install fiber at each street light. Our study focuses on providing a framework to help operator’s make the correct decision for their backhaul architecture.

Terragraph Backhaul Assumptions

• For each LOS link, we assume a capacity in Mbps based on a link budget & pathloss

• In absence of performance information, we assume each link has a capacity of 2.5Gbps. This capacity assumption can easily be replaced with a link budget based on 60GHz propagation model (3GPP, n.d.) and performance curves from engineering teams.

• Downlink to uplink traffic ratio of 7:3 and scaling available capacity by 0.7

• Latency per hop = 1ms. Max acceptable multi-hop latency is 10ms.

Since Terragraph requires line of sight for backhaul, each additional site that is brought into the network needs to have LOS with a site connected to fiber either directly or indirectly. The number of hops needed to reach the fiber adds to the latency and contribute to cumulative traffic at each hop. This cumulative traffic on each hop limits the number of sites that can be daisy-chained together.

Due to the architecture of Terragraph network coverage will expand in a hub and spoke configuration from the existing sites with fiber. As the network expands the spokes will begin to overlap providing uniform coverage and a greater diversity of connectivity.

To understand the dynamics of this network, a sensitivity model is run by varying the demand per site from 100Mbps to 2.5Gbps and varying the site count from 181 to 2000. By running the model for a range of scenarios while keeping the availability of fiber the same, we can simulate network performance as it is loaded increasing data traffic and sites.

The figure above demonstrates that as site count increases the amount of demand that can be supported per site reduces. This is primarily due to the capacity limitation of each hop and the cumulative traffic that has to be carried as site count increases.

Next, the figure below shows the distribution of the number of hops required to reach a fiber PoP. From the chart is clear the number of hops will increase as the number of sites increase.

If we define a minimum target demand of 500Mbps per site, our network as shown above in figure 3 can support approximately only 300 sites. At 400 sites the number of hops required to reach a fiber PoP increases, due to which the demand per street light needs to be lowered to remain within the capacity available at each hop. In order to expand the network to 2,000 sites, we would need to significantly increase fiber PoPs beyond the original 181.

Terragraph Network Deployment

AT&T plans to grow its San Jose network to 1,000 sites over the next 2-3 years. For this case study, we assume that the operator will plan to support a minimum traffic demand of 500Mbps per site. Due to its architecture, the coverage of a Terragraph network will expand in a hub and spoke configuration from existing sites which have fiber PoPs. As these deployments expand the spokes will overlap to give a uniform coverage and greater diversity of connectivity.

The image below shows a subset of San Jose where the red stars represent locations of Fiber PoPs, the green dots are street lights and the Blue line is the LOS link between two street lights with a Distribution Node (DN) at each end.

Backhaul Economics

An operator has multiple options to meet its backhaul needs for a 5G small cell including dedicated fiber at each street light or a solution like Terragraph. For this analysis, we assume each Terragraph fiber PoP needs to support 5Gbps or enough capacity for 10 sites based on our demand assumption of 500Mbps. The fiber PoP used in our traditional backhaul architecture is dimensioned to support 0.5Gbps since it will only need to support a single site.

There are 3 main differences in these two approaches of deploying backhaul

1. Terragraph requires fewer fiber PoPs

2. Terragraph has an additional cost for distribution nodes (DN)

3. Terragraph shares its fiber PoP between multiple sites resulting in higher capacity requirement per fiber PoP.

In order to understand the network economics, we use a metric referred to as network expense which is simply depreciated capital expenditures + operating expenses. Network expense allows an operator to annualize their network cost to is ultimately the number that hits their income statement.

The figure on top shows a scaled network expense of these two scenarios where the traditional backhaul costs 1x in Year 2018. In the year 2018/2019 due to the cost of additional equipment and higher rate Fiber POP the network expense for Terragraph is higher. From year 2020 onwards as the network grows so do the savings resulting in lower network expense over a 5-year period and beyond.

Conclusion

High-density urban deployments of small cell networks will require extensive backhaul. The cost of adding wired backhaul drops to thousands of sites will prove too costly. This study shows that Terragraph offers a lower network expense for a small cell network with the added benefit of being able to move unused capacity from one end of the network to an area where it is needed most, enabled by its mesh-based architecture. This results in better utilization of fiber assets and lowers the effective $/Mbps. When it comes to larger network deployments, Terragraph, has a clear advantage in terms of economics and ease of deployment. Operators planning for high density small cell deployment should seriously consider Terragraph as an option before any 5G rollouts begin. Terragraph offers scale, ease of deployment, and capacity requirements in a cost effective solution to meet the needs of high-density greenfield deployments that we will see with 5G. Solving backhaul connectivity is a critical step to deploying 5G networks in a scalable manner.

For additional details, you can view the white paper online here.

The opinions expressed in this case study are based on publicly available information and do not claim to represent any companies views or strategy.