Backhaul Planning for the Two Phases of Small Cell Rollout: 1 Reactive Tactics and 2 Proactive Strategy

Transcription

1 Backhaul Planning for the Two Phases of Small Cell Rollout: 1 Reactive Tactics and 2 Proactive Strategy February 2015

2 Backhaul Planning For HetNets 2 Summary Small cell backhaul is an integral part of the HetNet and needs to be jointly planned in conjunction with the radio access network (RAN) components. As small cell rollouts gather pace, the industry is gradually getting to grips with what is important and how different ecosystem players will need to collaborate in order to make small cells profitable for all. Our conversations with leading experts have uncovered emerging areas of consensus around the following themes: Design for app coverage. Demand and performance must be quantified in terms of the number of simultaneous users of a given app at a given location. App demand can be both indoors and outdoors and served by small cells in either location. Site and backhaul selection must factor in performance, cost and lead time. Time to Revenue (TTR) will initially be as important as Total Cost of Ownership (TCO), as it is critical in the reactive early phase of rollout described below. Closer collaboration is needed between operators and providers of real estate and backhaul. This will require sharing of commercially sensitive data such as search rings using secure formats Open interfaces are needed to join the many pieces of the HetNet planning puzzle. Legacy has left operators with many different tools for planning and monitoring their networks. Those with open interfaces will be better placed to foster collaboration and interwork with specialist tools and data. New backhaul models need to be incorporated as layers in HetNet planning: o Fibre is preferable from a performance perspective, but high TCO and unpredictable TTR are a barrier. Providers need access to operator search rings to reduce uncertainty in their TCO and TTR Service Level Agreements (SLAs). o Microwave in its block licensed form will be a staple solution and like V-band (60GHz), will need 3D Line of Sight (LOS) prediction capability in the HetNet planning tool. o Non-LOS wireless is attractive, but interference and NLOS throughput reduction must be accurately modelled if performance is to be relied on. Small cell rollout is currently in a tactical start-up phase since macro upgrades and new spectrum are largely coping with the rate of rising demand. Small cells are being deployed reactively in small numbers to fix localised issues. Over time, the macro network will reach its upgrade limits, and wide area densification with small cells will become the more cost effective option. Small cell deployment strategies and the role of planning tools will differ in these two phases as follows: Phase 1: Reactive rollout of individual small cells to fix localised app performance not-spots Performance monitoring identifies areas of poor performance in terms of unserved app demand. Operators then have a window of opportunity to fix it before dissatisfied customers churn. HetNet planning tools generate small cell search rings for export in a secure form to real estate and backhaul providers. Rings incorporate HetNet interactions, small cell app coverage, propagation etc. TCO and TTR data will be gathered for candidate sites and their backhaul options. Performance modelling must factor in HetNet co-ordination as well as assessing potential backhaul bottlenecks. Candidate sites and their backhaul options are then scored to identify the lowest hanging fruit for operators which will be fast to deploy, significantly improve app coverage and be low cost. Phase 2: Strategic scale rollout of small cells in clusters for wider area densification Diminishing returns for macro upgrades mean they are no longer the most cost effective approach for uplifting app capacity. Wide scale densification with small cells will become necessary, and a wholesale approach to planning and deployment will have merit. A proactive approach will reduce reliance on agile deployment and low TTR, allowing a heavier weighting on performance and cost. Established relationships between operators and providers of real estate and backhaul will be evident in volume based master attach deals with SLAs including TCO and TTR. The roles of the HetNet planning tool will be: o Forecasting of unserved app demand to identify larger areas in need of densification.

3 Backhaul Planning For HetNets 3 o o Automated site identification based on master attach deals, avoiding delays and complexity of search ring based requests. Optimal selection of a subset of candidate sites and backhaul in a cluster. In this later proactive phase, more weighting can placed on performance and TCO and less on TTR The reactive phase is currently underway in many developed markets. The timing of the second phase will depend operators spectrum holdings and loading levels. We see that the initial tentative phase is a necessary training period in order to build the relationships and understanding needed for the larger scale operation that second phase will be. Likewise, sophisticated automation capabilities in the HetNet planning tool for phase two need to be built on mature methods for quantifying benefits, TCO and TTR of candidate sites and backhaul options, developed during the initial phase. We are currently watching the industry learning how to walk, before it can run. Contents 1 Backhaul Planning in the HetNet Era Emerging Themes for HetNet Backhaul Planning Designing for App Coverage Timing is Everything New Backhaul Technologies The Secret of the Search Rings Too Many Different Planning Tools Shifting from Reactive Monitoring to Proactive Planning Changing Needs of Automated Site and Backhaul Selection Closer Collaboration between Operators, Real Estate & Backhaul Providers Steps to Plan Small Cells and Their Backhaul to Enhance Application Coverage Identifying Areas of Unserved App Demand Overall App Demand Existing Network Performance Unserved App Demand Calculation of Search Rings Candidate Sites Backhaul Options for Each Site Fibre and Wireline Line of Sight Wireless Non Line of Sight Wireless Scoring Candidate Sites and Backhaul Options Benefits Costs: TCO and TTR Value Score Automated Site and Backhaul Selection Conclusions For Further Information... 20

4 Backhaul Planning For HetNets 4 1 Backhaul Planning in the HetNet Era The rise of mobile data usage is changing the shape of the networks needed to support it. In densely populated urban areas macrocells are being augmented with small cells, DAS, Carrier Wi-Fi and License Assisted Access which must all work together seamlessly in the Heterogeneous Network (HetNet). We have discussed these changes and the impact on RF planning in our partner paper "The Role and Benefits of RF and Performance Modelling Tools in the HetNet Era" 1. This concludes that a holistic planning approach is needed to model interactions between the different components of the HetNet: macro and small cells with interference co-ordination, indoor outdoor propagation, and the subject of this paper: Joint planning of the RAN and its backhaul. Planning tools will need to consider small cell backhaul as part of the HetNet, to be jointly optimised with the RAN 1 Small cells can be deployed at many more types of site than macrocells because they are compact, which in turn is because they are low power. This does however reduce their range and so they need to be located close to the users they will serve, down at street level. Whilst this initially appears to solve the problem of getting good RAN connectivity to mobile devices, all it really does is to push the problem onto the last mile backhaul connection. Recognised as one of the main barriers to small cell deployment 2, significant industry attention has been focussed on small cell backhaul over the past few years: Operator alliances have stated their requirement that cost needs to be significantly less without impacting connection quality 3. Equipment vendors have responded with a variety of new technologies, particularly in the wireless space 4, and backhaul SLA 5 standards have been augmented to incorporate the different challenges of connecting to small cells 6. Significant changes are afoot then, which require us to take a fresh look at the process of planning deployments of small cells and their backhaul. As urban small cell rollout starts to gather pace, we have talked with some of the industry s leading experts to capture current thinking and discover developing areas of consensus around best practices. Section 2 summarises the common themes that have emerged from these discussions. Section 3 integrates these ideas into a step-by-step planning process which scores and ranks small cell sites and their backhaul options. This gives a value score representing the benefits they bring balanced against their total cost of ownership (TCO) and time to revenue (TTR). The purpose of this is to identify the lowest handing fruit that operators should deploy first: the sites and backhaul that can be rapidly deployed at low cost and bring the largest benefits to the end user experience. In all this we see a central role for the HetNet planning tool, which can model interactions between macro and small cells, DAS, Wi-Fi, LTE-LAA and a range of backhaul technologies, as well as their TCO and TTR. 1 The Role and Benefits of RF and Performance Modelling Tools in the HetNet Era", Ranplan, Sep 2014, 2 "Market Drivers for Urban Small Cells", Small Cell Forum, Feb 2014, 3 Small Cell Backhaul Requirements, NGMN Alliance, Jun 2012, 4 "Backhaul Technologies for Small Cells", Small Cell Forum, Feb 2013, 5 SLA= Service Level Agreement 6 Mobile Backhaul Phase 2 Implementation Agreement MEF , Metro Ethernet Forum,

5 Backhaul Planning For HetNets 5 2 Emerging Themes for HetNet Backhaul Planning As urban small cell deployments get underway, operators, equipment vendors, real estate and backhaul providers are getting to grips with the key issues and challenges that must be factored into the planning process. The following sections identify the commonly reoccurring themes that came out of our recent conversations with some of the industry s leading experts. 2.1 Designing for App Coverage People use their mobile connections for a variety of different services, or apps, ranging from low data rate messaging to high rate video streaming, delay tolerant cloud syncing to delay sensitive interactive apps like gaming and calling. At home or in the office, the available mobile broadband capacity is often much greater than the demand, congestion is uncommon and so most services get whatever they need. In busy outdoor public spaces however, demand often exceeds supply, and the network needs to intelligently manage traffic related to different apps in order that they all remain usable. Networks need to understand the types of traffic they are carrying so that they can be treated appropriately. When planning networks, operators are now designing for app coverage the number of users that can have an acceptable experience of a given app at a given location. Although Mbps per cell or per network does give a general indication of the overall capacity being supplied, it does not guarantee that all users are having an acceptable quality of experience on their apps. By way of example, some operators are installing LTE small cells in urban areas specifically to increase availability of video streaming services. Other services - like messaging might better be served using a wide area macro site. This focus on a particular app reduces the range of site and backhaul options available as we shall see later. Figure 1 illustrates a simplified view of app coverage for a small cell: higher rate services like video will have a smaller coverage due to the higher data rates required. In practice, app coverage will be impacted by a variety of other factors which are discussed later under the section on search rings. Figure 1 App coverage depends on the App s performance requirements, amongst other factors Designing for app coverage requires us to consider demand on a per app basis too. App performance depends on all the links in the end-to-end (E2E) connection and could potentially be limited by the radio access network (RAN), the backhaul, the end server or anything in between. It s generally not tractable to model everything at the same time, so the most likely bottlenecks should be the priority. For small cells, these are the RAN and the last mile backhaul.

6 Backhaul Planning For HetNets Timing is Everything In comparing different sites and backhaul options, the time before the site goes live and starts generating revenue is an important criterion, and should be factored in when picking out the best sites to deploy first. Lead times for site acquisition, small cell and backhaul installation and provisioning all contribute to the Time To Revenue (TTR) 7. Long TTRs can adversely affect an operator s profitability in the following ways: Long or unpredictable delays in site acquisition or backhaul provisioning lead to organisational inefficiencies, with cost incurring personal and assets on-hold and inactive and not contributing to revenue generation. Failure to address poor app performance in a given area will eventually lead to churn where customers move to another operator. Figure 2 Short deployment lead times help operators reduce deployment costs and customer churn. Some operators have described a just-in-time business model where network performance is heavily monitored to identify areas of congestion. The ability to then rapidly deploy small cells to resolve issues in key areas ensures the operators avoid churn whilst minimising deployment of excess capacity. We see that this reactive approach is applicable in the current initial phase of tactical small scale rollout to fix specific congestion issues. It may however not cope when larger scale densification is needed across wider swathes of the network. Figure 5 and later sections discuss these different phases. 7 5 ways to expedite LTE site deployment, Commscope, Feb 2015,

7 Backhaul Planning For HetNets New Backhaul Technologies The new challenges presented by small cells, not to mention the anticipated sales volumes, have driven investment and innovation into backhaul technologies, many of them wireless. WiMAX radios with their dynamic adaptation capabilities have been re-purposed for Non Line-Of-Sight (NLOS) backhaul in sub 6GHz spectrum in both licensed, license exempt, paired and unpaired combinations. At the other end of the scale, V and E band solutions (60GHz and 70-80GHz, respectively) exploit the vast amounts of relatively unused bandwidth here. Traditional microwave backhaul frequencies in the middle (20-42GHz) are being offered in innovative ways too: Block licensing (as opposed to link licensing) enables the licensee to manage their own interference, improving time to revenue and enabling the more flexible multipoint topologies. Non-LOS propagation has also been demonstrated at microwave frequencies 13, and is in use in commercial networks today. A detailed comparison of the characteristics of these technologies is available from the Small Cell Forum in its backhaul technologies document 17 To gauge current interest in these different options, we informally surveyed a cross section of the industry (excluding the backhaul providers themselves) and show our findings in Figure 3. Technology interest was triaged into whether it was a) likely to be used, b) being considered, c) not of interest, and scores compiled. Whilst we do not claim this is rigorous scientific study, we do feel it provides an interesting perspective. Figure 3 Qualitative comparison of interest in the different urban small cell backhaul technologies. Source: Ranplan informal industry survey, January 2015 Survey results largely tally with our expectations, with the exception of the strong interest in microwave solutions. Microwave has not received as much fanfare as Fibre, V-band and NLOS over recent years. This might be because it has already been well proven as a reliable backhaul technology in fixed point-to-point links, and therefore its acceptance has been unceremonious. Satellite was included as a control and as expected is not being considered for urban backhaul. It is however widely used in Rural and Remote and temporary scenarios 8. Overall, the perceived view of the different small cell backhaul technologies is: Fibre is ideal, but only when is available at the site with reasonable TCO and low and predictable TTR which often is not the case. Wireless technologies can be deployed by the operator, and thus have lower or more predictable TTR. They need careful planning, which is already a core expertise of the mobile operator. o Microwave: 20-42GHz will be a staple solution in its block licensed multipoint form. o V-Band is of great interest due to the Gbps throughputs achievable, license except planning and compact units. Cost, range and precision pointing on LOS links are all challenges, but ones which can be managed by RF planning. o Operators are very attracted to the flexibility and low TTR of Non LOS, but accept that spectrum limits performance and scale of deployment. One particular TTR sensitive operator considered using NLOS as a stop-gap until the fibre arrived even in its unlicensed form. xdsl over copper is also attractive due to its wide availability and low cost driven by the fixed broadband market. Rapid bandwidth drop-off with range and asymmetric latency are challenges for the tight sync needed for HetNet co- ordination. 8 " Overview of Rural and Remote Small Cells ", Small Cell Forum, Mar 2015,

8 Backhaul Planning For HetNets The Secret of the Search Rings Search rings indicate areas in which a new cell site will provide a valuable enhancement to the service level traditionally adding coverage or enhancing capacity. Since they essentially identify the areas of poor performance in an operator s network, they are also a closely guarded secret, and would not want to fall into the hands of competitors. On the other hand, fibre backhaul providers are expressing frustration that operators are not helping them to refine their service offerings by identifying the key areas of interest. The concept of search rings is not a new one, and they have been much used in the build out of macro mobile coverage for many years. They are an established way for operators to instruct site acquisition agents to identify potential real estate 9. Search rings can be applied to the small cells and app coverage too, but need to be adjusted slightly to take account of a range of HetNet era factors as illustrated in Figure 4. Figure 4 Search rings in the HetNet era must take account of desired app performance for the user cluster, as well as existing coverage level HetNet co-ordination required HetNet era search rings for small cell sites should take into account: The size, user density and app mix of demand that must be served The app coverage range of the type of small cell to be installed, which in turn depends on: o Power and antenna gain: whether beamforming or directive antennas used. o Local propagation environment: buildings, urban canyons etc. o Carrier bandwidth(s) for the small cell. o Whether co-channel with a co-covering macro site, and if so, the type and degree of interference coordination used (eicic, CoMP etc.) The user density in the cluster and the small cell capacity may be limiting factors, although it is not yet established whether these are best factored into search rings, or left for a later stage of planning. Taking all these factors into account will require the use of a HetNet planning tool. The end result will looks more like contour lines than rings, with the different lines representing apps rather than altitude. Search rings should be exported in a secure format for the intended reader only. 9 How Specific Cell Tower Sites are Chosen,

9 Backhaul Planning For HetNets Too Many Different Planning Tools The legacy of having used different technologies over the years, as well as objectives evolving from outdoor voice coverage to indoor and outdoor app coverage today, have left many operators with a plethora of different planning tools. Whilst each may do their own job well, interactions between indoor and outdoor, RAN and backhaul cannot be taken into account, and furthermore the complexity limits the scalability. The overall approach needs rationalising into separable parts which each fit together like a jigsaw using open interfaces. We have discussed this in our partner paper on planning needs for the HetNet era 1. In section 3 we illustrate the information flows between different parts of the HetNet planning process, and which can be outsourced to a third party or a legacy or specialist tool. 2.6 Shifting from Reactive Monitoring to Proactive Planning The just-in-time reactive approach to small cell deployment described earlier appears appropriate for the current early stage of small cell rollout to fix specific network issues. This approach is driven by feedback from performance monitoring systems which identifies symptoms of congestion and poor app performance. This then gives the operator a window of opportunity to fix it before dissatisfied customers churn. An agile small cell deployment strategy is then needed to be able to rapidly address issues before it s too late. This underlines the importance of TTR for operators using this type of deployment approach. Whilst reactive deployment seems reasonable at a time where the bulk of capacity uplift comes from macro upgrades and small cells are deployed individually for firefighting localised issues, it is unlikely to scale to wide area densification. When this comes, a more proactive wholesale approach will be needed where performance uplift is planned over larger areas. HetNet planning tools will identify optimal sets of small cell sites and backhaul that achieve a good performance / cost balance, and have reduced sensitivity to TTR. The role of monitoring will then be to validate that the planned performance is achieved on the ground. Figure 5 illustrates the shift in emphasis between these two phases or rollout. Figure 5 Shifting emphasis on time, cost and performance over the two phases of small cell rollout 2.7 Changing Needs of Automated Site and Backhaul Selection Automated site selection helps the operator understand which are the best small cell sites to deploy. We have seen a re-occurring theme that current planning approaches will change as rollout scales up. This theme applies also the type of optimisation needed for small cell site and backhaul selection: In the current reactive tactical phase, operators are targeting specific problem areas of congestion or poor app performance with one or two carefully placed small cell sites. Operators wish to identify sites in a given city representing the lowest hanging fruit to deploy first. These are the sites that score highly in terms of the largest benefit of improved app coverage for the lowest cost to deploy and fastest time to revenue. As rollout scales and operators are more widely densifying for capacity uplift, multiple small cells will need to be deployed across a wider area. The role of automation will then be to pick out a subset of candidate sites and backhaul options in larger clusters, with a given performance/cost target and reduced weighting on TTR.

10 Backhaul Planning For HetNets Closer Collaboration between Operators, Real Estate & Backhaul Providers Small cell deployment requires three key groups to collaborate and divide up the value chain in a way that is profitable and sustainable for all. Doing so requires the development of trust to share commercially sensitive information between parties. The coverage and capacity performance offered by a mobile network reveals much about its operator s strengths and weaknesses. Equally, a backhaul or real estate provider s offerings and pricing plan depends on who s asking. During the early days of ecosystem development, players keep their cards close to their chest, as laying them on the table would leave them wide open for competitors to take advantage. As the small cell ecosystem matures, players establish trusting relationships which allows for closer collaboration and mutually beneficial win-win relationships as illustrated in Figure 6. Figure 6 Maturation of the small cell ecosystem leads to closer collaboration and mutually beneficial relationships Our vision for the HetNet planning tool is that it will help foster collaboration between the three parties needed to make HetNets a reality. It provides import and export formats to enable these parties to securely share sensitive data and enable operators to design better networks for the benefit of all. Later we elaborate on how such a planning tool can create the search rings needed by real estate and backhaul providers to focus their offers on areas of most interest, improving the quality of their TCO and TTR estimates.

11 Backhaul Planning For HetNets 11 3 Steps to Plan Small Cells and Their Backhaul to Enhance Application Coverage This section takes the various needs identified in the previous section and synthesises them into a step by step process. This reveals the various input and output data, the parts that can be performed by the planning tool and which of these could potentially be outsourced or imported. Open yet secure interfaces will be needed to maximise flexibility and interaction with third parties. Figure 7 shows a flow diagram for this process and Figure 8 illustrates examples of data sets that would be involved. Each step is elaborated in the subsequent sections. Figure 7 Overview of process to identify sites that will most rapidly and cost effectively improve app coverage where needed

12 Backhaul Planning For HetNets 12 Figure 8 Types of data needed to compare the benefits, TCO and TTR of different small cell sites and their backhaul options

13 Backhaul Planning For HetNets Identifying Areas of Unserved App Demand Deployment of additional cell sites must be driven by demand that is not already served. Operators have to react to sudden localised surges in demand that their networks cannot support. Ideally though they will look ahead with a demand forecast which predicts evolving uptake and usage of services, and have a capacity plan in place to accommodate it. The following sections describe how app demand is forecast and the existing network capacity removed to leave hotspots of unserved demand where small cells can be deployed to good effect Overall App Demand Demand needs to be specified in terms of the maximum number of users simultaneously requiring a given app at a given location, since this is ultimately what the operator is designing the network to deliver. Although it is simpler to consider demand density in terms of Mbps/m 2 this does not differentiate between say many high value internet shoppers or a file sharing bandwidth hog. Furthermore, it does not take into operators ability to manage traffic during times of heavy use using E2E QoS and so on. Services like voice and video which are become un-usable with poor connectivity take priority over delay tolerant services like cloud storage or syncing. Voice and video codec rates can also be adapted to ensure the experience remains acceptable under all load conditions. With a mix of services in use, they can be grouped into a number of different levels for bandwidth demand. In Figure 8 we illustrate two levels: Video, which requires moderately high rates to sustain a good experience, and what we call HTTP, representing services that can tolerate lower rates and higher delays. Our example plots in Figure 8 show a general background level of demand with a number of higher density user clusters of various shapes and sizes. The similarity between the Video and HTTP plots is driven by high user density in areas such as station exits, cafés, and so on. Some have the view that LTE small cells will be deployed primarily to serve high rate users in dense clusters. HetNet planning tools need to be able to explicitly model app demand: clusters of users in specific locations using specific applications and later test whether they have sufficient connectivity to ensure their experience is acceptable. Traffic models can be used to represent the different types of app Existing Network Performance The existing network s app coverage must be factored in to identify potential areas for enhancement. It may be possible to export this from a macrocell planning tool. Ideally however, the macro network would be explicitly modelled in the HetNet planning tool, ensuring interference co-ordination is correctly represented as well as app QoS prioritisation and layer management. These functions allow operators to bias particular services onto or away from the small cell or other layers. This both balances loading across the layers as well as matching service characteristics to the layer. For example macro cells are better for wide area mobility, whereas small cells are more likely to offer higher peak rates as fewer users are sharing and signal qualities are generally higher Unserved App Demand The resulting plot of unserved demand describes the number of users at a given location wishing to use a particular app, but would not be served by the existing infrastructure. In Figure 8 we show that the macro network has served the general background level of demand, leaving only isolated hotspot peaks which are ideally suited to small cells. Unserved demand is not necessarily restricted to outdoor areas, and the HetNet planning tool should be able to model indoor and outdoor demand and sites, and propagation in between. 10 Small Cell Backhaul Requirements, NGMN Alliance, Jun 2012,

14 Backhaul Planning For HetNets Calculation of Search Rings Sharing of search rings may greatly improve collaboration with real estate and backhaul providers, as it enables them narrow down their search for candidate options, reducing margins for pricing and lead time uncertainly. If site and backhaul option cost and timing data are already available, then this step can be avoided. The search ring essentially represents the locations in which a deployed small cell would serve a worthwhile amount of demand, i.e., that would be profitable for the operator. As discussed earlier, evaluation of search rings requires a planning tool to take into account a number of factors: the performance of the given app, the app coverage range of the small cell which in turn depends on the local topography, interference and co-ordination with other co-channel sites, etc. Although described as rings, in practice they would be contours of app coverage benefit which morph around buildings. The search rings for HTTP are larger than that for video in Figure 8 because the required performance is lower and so a small cell can achieved this at greater range. Search rings are commercially sensitive and should be exported in a secure format for the intended reader only. 3.3 Candidate Sites Candidate sites may come from real estate owners or acquisition agents having been provided with search rings, or might be automatically identified using a modelling tool. In addition to latitude, longitude and height, it will also be necessary to know pricing and lead time information to combine with backhaul option data to evaluate TCO and TTR for later site + backhaul scoring and ranking. Neutral host providers may combine site rental, power, backhaul connectivity and installation and other services into a single figure. The HetNet planning tool needs to be capture and combine TCO and TTR data in different ways to represent these different business models potentially co-existing side by side in the same operator network. 3.4 Backhaul Options for Each Site Figure 8 shows three basic types of small cell backhaul: fibre, Line of Sight (LOS) wireless and Non-Line of Sight (NLOS) wireless. In practice the HetNet tool has dedicated models for each wireless frequencies and their specific propagation characteristics. A framework which enables articulation of different types of fibre offering is also needed. We provide example of one type of offering which captures TCO and TTR for different locations. In Figure 8 we use TCO levels 1-4 to broadly represent relative costs between different backhaul options. In practice, a number of CAPEX and OPEX figures will need to be input in the local currency. In cases where a multipoint wireless hub is deployed, this might be considered as an initial CAPEX for the first link, or amortised over all links. When a backhaul hub is deployed at a macrocell site, there may be an additional cost to upgrade its fibre backhaul.

15 Backhaul Planning For HetNets Fibre and Wireline Extensive fibre networks run throughout densely populated areas, and can be offered in a variety of forms including backhaul suitable Metro Ethernet Services as described in MEF22.1 and amendment for small cells 11. Fibre may also be offered in wave division multiplexed (WDM) formats, needed to transport the very high rate Common Public Radio Interface (CPRI) protocol for Centralised RAN deployments 12 Figure 8 shows a typical managed fibre offering three levels of pricing and availability as follows: Existing cabinets or points of presence (PoPs) will be the lowest price and shortest lead time. Green zones are areas close to existing routes and PoPs. Connections here have medium price and lead time. Beyond green zones, costs and lead times are much higher and also harder to predict. Performance wise, fibre technology is more than capable of meeting the requirements for a backhaul service. Where many cells are chained together, or where copper or a bandwidth limited service is used, the backhaul could become the limiting link in the end to end connection. Bandwidth will need to be input for all backhaul technologies to ensure any limit is reflected in the app performance Line of Sight Wireless Microwave and V-band (60GHz) are anticipated to be widely used for small cell backhaul. Although these typically require a Line of Sight link, one vendor has even shown that 28GHz microwave has sufficient link budget to operate in Non Line of Sight for short range small cell scenarios 13, and claims several such links in commercial operation. The HetNet planning tool can use the same 3D topographical data used for RAN coverage to predict Line of Sight coverage from a chosen hub site for rooftop-down type deployments. It can equally predict which of the candidate sites could see each other at street level, to allow chaining of a number of sites back to one with a fibre connection. This topology is proposed by Virgin Media for their hosted small cell service 14,15. Both roof-street and street-street LOS links are illustrated in Figure 8. Block licensed microwave spectrum is preferable to operators as it allows them to manage frequency planning inhouse, rather than including the regulator in the loop. A HetNet planning tool modelling 3D buildings in the urban environment will enable operators to re-use block licensed microwave frequencies much more tightly than the terrain only based predictions used for conventional point-point frequency allocation. Line of Sight modelling for small cells factors in Fresnel zone clearance 16, as well as additional margins to allow for inaccuracies in 3D building data and site co-ordinates. Similar techniques can be applied to E-band backhaul modelling, noting that: E-band is license exempt so anyone can transmit anywhere (within EIRP limits). Interference from such sources is unpredictable and is therefore difficult to model. High atmospheric absorption, very narrow beam antennas and street level deployment reduce the likelihood of interference being a problem. Ranges of a few hundred meters could be a limiting factor for E-band small cell backhaul. V-band has longer ranges and even higher capacities than E-band, and can be modelled in a similar way. Each frequency band has unique characteristics which dominate in propagation predictions, and appropriate validated models must be used. 11 Mobile Backhaul Phase 2 Implementation Agreement, Metro Ethernet Forum, 12 CPRI transport in C-RAN Networks, K. Murphy, Ericsson, Dec 2014, 13 Non-line-of-sight microwave backhaul for small cells, J Hansryd et al, Ericsson, Feb 2013, 14 Virgin Media's offering on SCaaS, 3G4G blog, Fresnel Zone, retrieved Feb 2015,

16 Backhaul Planning For HetNets 16 Line of Sight microwave links have channel sizes between 14MHz 112MHz and typically operate at high SINRs, and so achieved throughputs into the high 100s Mbps to over 1Gbps 17, which is more than adequate for small cell backhaul. Multipoint configurations can be used to increase deployment flexibility and exploit gains from statistical multiplexing. As with fibre, bandwidth could still become a limiting factor where multiple small cells are backhauled in a multipoint fashion, or daisy chained. As always, backhaul bandwidth should be factored into the E2E app performance calculation Non Line of Sight Wireless NLOS wireless backhaul works in a similar way to the RAN access link itself it can refract around corners, penetrate through or reflect off of walls. This brings great deployment flexibility, but generally offers lower bandwidths than LOS links at higher frequencies where more spectrum is available. RF modelling for Non Line of Sight backhaul is essential to ensure that sufficient performance can be achieved. As with the RAN itself, the limited and highly valuable spectrum must be re-used as much as possible to maximise overall backhaul capacity. This means that interference is highly likely for any sizeable deployment with more than one hub, and must be taken into account. The more torturous the propagation path, the lower the received signal power will be relative to noise and interference, and thus the lower the link throughput. Multipoint topologies are possible, but the degree of sharing may be limited by amount of spectrum and the NLOS throughput reduction. TCO and TTR modelling for NLOS is similar to LOS wireless: Both ends of the link must be accounted for, and ant transport capacity upgrades needed for the backhaul hub itself factored in. Note that for simplicity, we have not included the NLOS backhaul option in value scoring example in Figure Scoring Candidate Sites and Backhaul Options The objective of the planning process described thus far is to identify the lowest hanging fruit sites to deploy first. These are the sites that can quickly be up and running to capture a large amount of previously unserved app coverage, at low cost. These sites are of high value to the operator as benefits are large and costs are small. Value = benefits - costs Value = Increase in satisfied app users - TCO TTR In our planning example, we can now quantify benefits and costs for each of the nine candidate sites, and combine these to arrive at a value score, as illustrated in Figure "Backhaul Technologies for Small Cells Small Cell Forum, Feb 2013,

17 Backhaul Planning For HetNets 17 Figure 9 Quantifying benefits of costs of candidate sites and their backhaul options to give a value score Benefits We measure the number of app users that would be within the app coverage area of a small cell at the candidate site. The above figure shows app coverage as a circle for simplicity. In reality it would depend on local buildings and obstacles and propagation around them, as well as interference and co-ordination with the macro and other small cells. The capacity of the small cell may limit the number of app users that can be served, and should be factored in. We show separate benefits scores for the video and HTTP services, and see the latter scores more highly due to the lower per user app throughput requirement. However, operators may wish to weight apps differently when considering the benefits. The weight row in this example in the table shows video users as five more important than HTTP. An operator deploying small cells only to enhance video might set the HTTP weight to zero. Setting benefit weights to represent the Mbps for each app user, would mean the overall benefit score would be the total number of Mbps within the catchment area of a given small cell Costs: TCO and TTR We consider both TCO and TTR scores for fibre and LOS backhaul options. We have omitted NLOS options for simplicity. TCO is expressed as a level between 1 and 4, as described in the table within Figure 8, which also gives TTRs for the different options in days. The values given are for illustration only, and will likely vary significantly from market to market. Detailed cost modelling figures for small cells and backhaul can be found in the Small Cell Forum s Urban Business Case paper 18. Note the weightings for TCO with respect to TTR are to make the numbers of comparable magnitude in the valuation score. In practice the actual weights will depend on operator preferences and the units used for TCO and TTR. The agile reactive deployment phase described earlier requires a higher weight for TTR than the later proactive phase. 18 "Business case for Urban Small Cells", Small Cell Forum, Feb 2014,

18 Backhaul Planning For HetNets 18 Note that sites 1 and 4 have a potential LOS to site 3, and so could benefit from its fibre connection. The LOS TCO and TTR for these sites assume that site 3 has already been deployed, paid and waited for. Whilst this may make them look attractive, we should further note that once site 3 has been deployed - the unserved app demand available to these nearby sites will be significantly reduced. A useful option for HetNet planning tool would be to assume that the higher ranked sites have been deployed, and their app coverage subtracted from the unserved demand Value Score The value score is the weighted sum of benefits and costs for each candidate site and backhaul option, normalised to a scale. Our method picks out sites 3, 6 and 9 as having high value scores and thus would be the first choices for the operator. Sites 2 and 8 both have low scores simply because they capture little if any unserved app demand. As discussed, site 4 shows an artificially high score, because we have not taken account here the app demand that would already have been captured by site 3. A similar value scoring and ranking method was used by Real Wireless 19 to assess a large portfolio of tens of thousands of real estate assets of different types for potential small cell deployment. The top few hundred sites are shown in Figure 10. This not only picks out the best sites to deploy first, but the colouring also gives a broad indication of the types of sites. Figure 10 Example Ranking candidate small cell sites according to operator perceived value Source: Real Wireless Automated Site and Backhaul Selection The process we have described is aimed at early small scale deployments, where small cells are being tentatively deployed in small numbers to relieve specific areas of congestion and removing the no-go areas for people wanting to use demanding apps and services. In this phase, operators just want to identify the best, cheapest and fastest places to deploy small cells as shown here. For urban outdoor metro cells, this early stage is happening now and is likely to continue for another year or two, depending on the region 20. As the new LTE spectrum is deployed and macrocell upgrades are used up, there will follow a more general need for wider area densification. It will then become more cost effective to deploy multiple cells at a time across wider areas. The HetNet planning tool will need to identify an optimal set of sites and backhaul to meet this larger requirement. Such an approach is demonstrated in our study on outdoor LTE small cell deployment on lampposts Small Cells on the Right Sites when Budgets are Tight or The hunt for the golden lamppost, Simon Saunders Real Wireless, Small Cells World Summit "Small Cells Market Status", Small Cell Forum, Mar 2015, 21 Outdoor LTE Small Cell Deployment on Lampposts: A Paris study, Ranplan, Jun 2014,

19 Backhaul Planning For HetNets 19 4 Conclusions We have considered the planning process for urban public access small cells and their backhaul, when deployed as part of a heterogeneous network (HetNet). Conversations with leading experts have uncovered areas of emerging consensus around a number of key issues: 1. Design for app coverage. Network demand and performance must be quantified in terms of the number of users of any given app at a given location. 2. Time to Revenue (TTR) will initially be as important as Total Cost of Ownership (TCO), as it is critical in the reactive early phase of rollout described below. 3. Closer collaboration is needed between operators real estate and backhaul providers this requires sharing of commercially data such as search rings, which must therefore be in a secure form Much of the discussion around small cell technologies and processes to date has focussed on the end game of what might ultimately be the ideal approach. However, we now see the industry turning to look at what is practical today based on real world constraints, legacy and ecosystem maturity. Small cell rollout is currently in a tactical start-up phase since macro upgrades and new spectrum are currently coping with the rate of rising demand. Small cells are being deployed in small numbers to fix localised issues. Over time, the macro network will reached its limits, and wide area densification with small cells will become the more cost effective option. Small cell deployment strategies and the role of planning tools will differ in these two phases as follows: Phase 1: Reactive rollout of individual small cells to fix localised app performance not-spots. Macro upgrades with additional spectrum and antennas will shoulder the bulk of demand increases. Heavy use of performance monitoring to identify congestion hotspots in need of fixing. Agile small cell deployment required, underlining the importance of time to revenue in order to avoid customer churn. HetNet planning tools perform the following tasks: o Quantify problem areas in terms of locations of users with poor or no app coverage. o Generate search rings for export in a secure form to real estate and backhaul providers enabling them to reduce uncertainty and margins on pricing and TTR (Time to Revenue) commitments. o Capture cost and performance data for candidate sites and their backhaul options. o Assess performance and cost of backhaul options, including 3D modelling of LOS and NLOS wireless links. o Score and rank candidate sites and backhaul options identifying for operators the lowest hanging fruit which will be fast to deploy, significantly improve app performance, at low cost Phase 2: Strategic scale rollout of small cells in clusters for wider area densification Diminishing returns for macro upgrades mean they are no longer the most cost effective approach for capacity uplift. Wide scale densification with small cells will become necessary, and a wholesale approach to planning and deployment will have merit. A proactive approach will reduce reliance on agile deployment with low TTR. This allows more weighting on performance and cost. Relationships between operators, real estate and backhaul providers will be established. A better mutual understanding will help reduce uncertainties on both sides for more cost efficient win-win collaboration. Master attach deals with SLAs can be agreed based on higher volumes. Proactive planning using a HetNet planning tool: o Forecasts of unserved demand will identify larger areas in need of densification o Master attach deals with providers will facilitate automated site identification with associated TCO (Total Cost of Ownership) and TTR data, avoiding need for search ring based requests. o Tools automate planning of small cell clusters and backhaul, optimised against performance and TCO targets and lower weighting on TTR

20 Backhaul Planning For HetNets 20 5 For Further Information This paper shares Ranplan s vision on the roles of HetNet planning tools, both now and in the future. To find out more about our own HetNet planning tool, ibuildnet, its current capabilities and feature roadmap, please contact us. We d like to hear your views too. info@ranplan.co.uk Ranplan Wireless Network Design Hart House Business Centre Kimpton Road Luton, LU2 0SX United Kingdom T: +44 (0) F: +44 (0)