Massively densified networks Why we need them and how we can build them Monica Paolini, Senza Fili
We thank these companies for sponsoring this report: In collaboration with
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |1| Table of contents
I. Report: Moving to dense, pervasive networks 3 II. Vendor profiles and interviews 29 1. Densification is more than small cells or DAS or C-RAN. From Anritsu 30 atomic to pervasive networks 4 Ascom Network Testing 37 2. The emerging RAN taxonomy. Antenna coverage and baseband CommScope 44 separation 8 InterDigital 52 3. The big question: Indoor / Outdoor 10 Kathrein 59 4. Drivers: Coverage / Capacity 11 Rohde & Schwarz 66 5. Performance: Capacity / Latency 12 Samsung Networks 74 6. Architecture: Small cells / DAS 13 SOLiD 81 8. Network: Distributed / Centralized 15 SpiderCloud Wireless 88 9. Technology: Cellular / Wi-Fi 17 III. Operators’ interviews 95 10. Unlicensed spectrum: LTE / Wi-Fi 18 11. Spectrum: Sub-6 GHz / Millimeter wave 20 BT 96 12. Interference: Co-channel / Separate channel 21 DOCOMO Innovations 104 13. Backhaul: Fiber / Wireless 22 Carolina Panthers 110 14. Fronthaul: CPRI / Xhaul 23 Enterprise, anonymous 116 15. Access point density: High / Low 25 Glossary 120 16. Business model: Single operator / Shared deployment 26 References 122 17. Concluding thoughts. The role of operators in a densified, pervasive network 28
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |2| I. Report: Moving to dense, pervasive networks
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |3| 1. Densification is more than small cells or DAS or C-RAN. From atomic to pervasive networks
We often equate densification with the small-cell deployments, usually in conjunction with Wi-Fi and DAS, that increase wireless network capacity to accommodate traffic growth. While small-cell deployments are certainly a central element in the densification process, densification itself is becoming the catalyst for a much deeper evolution in wireless networks, which is not limited to small-cell deployments, and which is enabled by demand drivers and new technologies but goes beyond the contribution that each of them separately brings.
Today’s networks have an atomic, discrete architecture in which cells are the edge access elements and are all connected to a distinct, common core. In the initial stages of densification, operators increase the number and density of these elements with a surgically precise addition of small cells, DAS or carrier Wi-Fi elements, but the architecture remains fundamentally the same. As this process intensifies, these atomic networks start to change into what we call pervasive networks in this report. Others (notably among them, I Chih-Lin at China Mobile) have called them user-centric networks, as opposed to the traditional network- centric network.
Pervasive networks are distributed. Edge elements of the network, such as small cells or DAS, get closer to users and devices. And devices themselves can become part of the access network itself, with device-to-device connectivity.
With C-RAN and, more generally, virtualization, the cell as the fundamental self- contained element in the RAN ceases to exist. It is replaced by a multi-layer, multi- band set of antennas connected to a remote baseband. Devices within this model can connect to more than one antenna and, in a mobility scenario, switch from one antenna to the next without having to do a handoff, because the cell ID remains the same. This is the evolution model put forward by China Mobile’s no-more-cells approach and DOCOMO’s phantom cells.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |4| Demand is the main driver to densification. Traffic growth on wireless networks continues unabated as subscribers conduct more activities and a wider variety of them, over more mobile devices, and as IoT spreads. In turn, the increased coverage and capacity of densified wireless networks facilitate subscriber and IoT device access, and this drives a further increase in demand. Increasing network capacity to meet demand has become financially unsustainable in today’s atomic networks. Pervasive networks allow a more efficient use of network resources that will enable operators to provide the capacity and performance needed cost-efficiently.
Outdoor small cells were the first solution to address the need for densification as a complement to Wi-Fi offload. Small cells can be deployed in both 3G and 4G networks, and can be combined with Wi-Fi, sharing spectrum with the macro layer. As operators started to test small cells and plan for deployments, though, they realized that moving to large-scale deployments of small cells required substantial operational and financial effort.
Over the last few years, the entire wireless ecosystem has been working to find business models and technological solutions that meet the operators’ performance and cost requirements. The rest of the report will discuss the advancements in this area and how they relate to pervasive networks in detail. For now, we can call out 5G and virtualization as the crucial technology enablers in the transition to pervasive networks. The two technologies combine the performance improvement, the cost effectiveness, and the flexibility that operators need to meet the growth in demand in their networks.
The transition from atomic to pervasive networks has a major impact on wireless networks – from technology, performance, usage model, and financial perspectives, as described in the table below. The evolution is not confined to wireless technology or network architecture. It affects the entire ecosystem, including subscribers, enterprise, and third-party players, as well as business, ownership, and operational models.
Densification is necessary for wireless networks to meet demand, but many changes, are necessary to achieve the capacity and performance goals, and they will eventually transform today’s atomic networks into pervasive networks.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |5|
Atomic networks Pervasive networks
Network model Network-centric: subscriber adapts to the network (e.g., goes to the User-centric: network adapts to subscriber demand (e.g., ultra-dense window to make a phone call). wireless infrastructure in stadiums). RAN Discrete elements: cells (antenna and baseband). No-more-cells, phantom-cells approach, with antennas as access points in a multi-layer topology, connected to a remote baseband. UE-RAN connection One-to-one connection from the UE to the cell. UEs can connect to multiple antennas, use multiple bands. Handoffs required for the UE to move association from one cell to Flexible modes of connectivity coexist: dual connectivity, device-to- the next. device connectivity, Wi-Fi offload. Subscriber can establish multiple concurrent connections: multiple devices (including non-SIM and IoT devices) on the same plan. Distinction between RAN elements and devices is less sharp because devices connect to each other and act as access points to the RAN. User and control User and data planes allocated to each access channel (e.g., sector). Control plane can manage traffic for multiple access channels, so planes some access channels do not require a separate control plane (e.g., LTE in unlicensed bands, LWA, mmW). Short-range mobility can be managed without handoffs. Densification targets Coverage in the wide area, capacity in high-traffic areas, with most of Vertical capacity increase and coverage extension driven by location- the RAN infrastructure in outdoor locations and large venues (e.g., specific traffic or service requirements (e.g., service tied to a venue; stadiums). IoT service). Layers Single macro-layer, possibly with limited small-cell hotspot Multi-layer networks, with extensive indoor and outdoor coverage deployments, and with Wi-Fi offload. with small cells, DAS or femto cells. Spectrum Cellular frequencies below 3 GHz. Wider range of higher-frequency bands (3.5 GHz, 5 GHz, mmW), with the inclusion of unlicensed spectrum. Core/RAN Separate location and equipment, with RAN equipment located at Boundaries less strict, with RAN becoming virtualized and the edge and core equipment in centralized locations. centralized, and some wireless core functionality moving to the edge (e.g., MEC, CORD). Location of function (distributed versus centralized architectures) is a strategic decision.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |6| Atomic networks Pervasive networks
Testing, monitoring, Testing and monitoring based on network KPIs and historical data. QoE metrics based on performance of UEs are tied to network KPIs optimization Limited optimization functionality. to test, monitor and optimize networks in real time.
Performance yardstick Capacity per RAN element. Capacity density (e.g., per sq km) and latency.
Traffic management Maximize throughput. Real-time traffic management, at the application or service level. Capacity determines service availability. Network slicing used to extract more value from network resources. RAN equipment Telecom assets (e.g., macro-cellular towers, building rooftops), RAN equipment gets closer to subscribers and devices – closer to the location mostly in outdoor locations. ground and indoors.
Network ownership Operator owns network, often leasing space on cell tower or other Venue owners increasingly pay for infrastructure, even though they assets. Limited network sharing. do not (and choose not to) operate the network. Multi-operator, neutral-host model, in which some network elements (e.g., backhaul) are shared among operators. Control Operators control end-to-end network. Operators retain control of the RAN, but other players – venue owners, residential users, neutral hosts and system integrator – get more visibility into the networks and have a stronger role in determining how the network resources they paid for are being used.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |7| 2. The emerging RAN taxonomy. Antenna coverage and baseband separation
Densification dictates an evolution of the RAN that involves all its elements. For operators, densification for capacity purposes starts with the macro networks. When operators need more capacity in an area, they typically first densify the macro network where possible, by adding cell sites, splitting sectors or adding new channels or bands. At some point this becomes financially too expensive or difficult (e.g., in environments where antenna density is already high, or where it is difficult to find new cell sites), and operators move densification to sub-layers – micro cells, small cells or femto cells. DAS deployments typically run in parallel, to address high-density venues such as stadiums or enterprises.
Distinguishing among different RAN elements, from macro cells to femto cells, including DAS, has become increasingly difficult. Many solutions are available to address specific environments, and do not neatly fit into any of the traditional RAN element groups. Rather, there is a continuum of solutions that are needed jointly, to address the varying requirements of different environments. This is a welcome evolution that testifies to the increased awareness of the multitude of environments where we need further densification, and the specific challenges that each presents.
At the same time, the trend toward C-RAN and, more generally, toward the virtualization of the RAN creates a second dimension by which to define RAN elements. The first dimension is the antenna coverage area, which decreases in the move from macro to femto cells, but without well-defined borders among the different element types. The new, second dimension is the location of the baseband. In a distributed, traditional RAN, the baseband is at the cell site at the edge. In a centralized network, the baseband is located remotely. There are different types of baseband separation, depending on the type of fronthaul used – and the type of functional split that defines different types of xhaul.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |8| Like macro cells, small cells can be distributed or centralized. Small cells and C-RAN are often presented as alternatives to each other, and small cells are seen as competing with DAS. However, within the frame of the densification process, small cells and DAS converge to a set of solutions with varying degrees of centralization – with DAS being always centralized, and small cells being either centralized or distributed. As a result, we end up with a continuum of solutions on both axes. This helps operators find a solution that is well-tailored to their needs and helps vendors create differentiation in the marketplace – but it also increases the complexity of the selection process.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |9| 3. The big question: Indoor / Outdoor
Mobile phones were initially developed to provide the wide-area connectivity that wireline phones could not provide. The first mobile phones used satellites for access and were car phones, and there was widespread doubt that mobile phones could be of any use in areas like Manhattan’s urban canyons or for more than short calls.
Today the situation is reversed. Most traffic – 80% or more in most markets – comes from indoor locations, and over 90% is data. Yet most of the RAN infrastructure, if we exclude Wi-Fi, is located outdoors; as a result, indoor coverage and capacity are more limited than outdoor. As wireless has become the default communication interface, the ability to provide the same level of service indoors and outdoors becomes a high-priority requirement for mobile operators – and one that, in most locations, cannot be met cost-effectively with only outdoor RAN infrastructure taking an outside-in approach.
Wi-Fi has been a boon for mobile operators: it transports the bulk of traffic from mobile devices, and most of that traffic is from indoor locations – public venues, enterprise locations, homes. While Wi-Fi continues to complement cellular access, mobile operators want to improve indoor cellular coverage, both to meet the demand from indoor subscribers and to relieve pressure on the macro network. Indoor traffic typically is more expensive to carry than outdoor traffic, because it uses less efficient modulation schemes and hence uses more network resources.
As a result, in recent years, operators have expanded their indoor coverage efforts. With the exception of some Asian countries, particularly Japan and Korea, mobile operators have been cautious about in-building networks, with the exception of large venues such as stadiums. Indoor coverage presents its own challenges, which are markedly different from those in outdoor deployments. The technology, the solutions and the business models are different, and they are evolving along with the relationship among venue owners, operators, and third-parties.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |10| 4. Drivers: Coverage / Capacity
The narrative of wireless network deployments is in many ways all about densification. Mobile network performance and capacity have improved tremendously, thanks to technological innovation and greater spectrum availability, and densification has been a key part of that improvement. Initially densification was used to address coverage holes, or establish more consistent coverage. With the growth in data usage, capacity requirements have become a major driver to densification – and densification has become a top priority for operators.
Increasingly, however, coverage and capacity have intertwined. Just being able to receive a signal at a given location is no longer sufficient for coverage there. Depending on the market, location and operator, the capacity required for adequate coverage varies, and it is becoming meaningless to define coverage without reference to minimum capacity requirements. As a result, a location that was deemed to have coverage in the past may no longer have basic coverage, and it becomes a new densification target in order to achieve sufficient capacity.
At the same time, boosting capacity in a hotspot may improve coverage in the surrounding area. An example is indoor infrastructure that addresses the demand created by indoor user offloads from the macro network serving the location; boosting that hotspot’s capacity also increases the capacity and coverage area of the macro network. In this case, the small cells or DAS in the indoor deployment not only increase the capacity density within their footprint, they also improve performance and coverage in the wider area.
The co-dependency of RAN elements within the same footprint in determining both capacity and coverage demonstrate the need – and benefits – of looking at densification efforts within all the layers of the RAN environment rather than on the RAN elements – e.g., small cells or DAS – that are directly involved.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |11|
5. Performance: Capacity / Latency
Capacity and coverage are the drivers to densification: they make it possible for operators to support the services that subscribers pay for. But capacity and coverage are no longer sufficient to make these subscribers happy. QoE, the quintessential – if somewhat elusive and difficult to quantify – measure of subscriber happiness, is increasingly determined not just by service availability (enabled by coverage and capacity), but by latency, as well as other transmission metrics such as jitter and packet loss.
Latency’s rise to prominence is due to the increased use of real-time data applications: streaming video and audio (such as YouTube, Spotify), voice (including VoLTE, OTT voice), entertainment and gaming (e.g., Pokémon). Some IoT applications – e.g., connected cars, safety, monitoring, medical, financial – have tight latency requirements, too.
With real-time applications, poor coverage, congestion and high latency affect QoE in comparable ways: they create a poor subscriber experience, with subscribers giving up on the application they want to use, or with application becoming unavailable. Common effects of high latency include delays on voice calls and games, and, with video, frozen streams, dropped frames, pixelization or long startup times.
As operators plan to increase coverage and capacity via densification, the ability to use low- and, as we move to 5G, ultra-low latency becomes a determinant when choosing the end-to-end network topology. The RAN plays a crucial role in latency, but so do backhaul/fronthaul, transport, core functionality, application and content processing and availability; they all need to be factored into network deployment and optimization for densified networks.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |12|
6. Architecture: Small cells / DAS
Small cells and DAS are frequently considered to be competing solutions taking opposed approaches to densification. Historically they have developed to address different requirements: small cells mostly for capacity, DAS mostly for coverage. They generally serve different environments, as well: small cells in outdoor and small-venue / residential locations, DAS for indoor environments and large outdoor venues (although DAS can also be deployed in other outdoor locations).
But DAS and small cells are quickly converging in a varied set of solutions that address the varied needs of venue owners and operators, and that combine features of both small cells and DAS. Both small cell and DAS vendors have increasingly expanded their portfolio to include both solutions or hybrid solutions (e.g., CommScope’s OneCell, Ericssons’s Dot and Huawei’s LampSite).
RAN virtualization pushes small cells even closer to DAS. Furthermore, DAS is a precursor of C-RAN. This is especially true of active DAS topologies that allow a higher level of control over the management of network resources.
An additional push for the convergence comes from the need to address medium- size venues – sometimes referred as the middleprise. Most in-building deployments target large venues because of those locations’ prominence and the dense demand there. Wireless performance during the Super Bowl, for instance, is tracked at an unparalleled depth. Locations like stadiums attract much attention and investment from venue owners and operators.
Smaller venues are a much larger market (e.g., in terms of footage), but are much more challenging to cover profitably, because that market is fragmented and, with some exceptions, smaller venues do not have high capacity requirements or high
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |13| revenue opportunities. Hybrid solutions and variations on the established DAS and small cell solutions – especially when coupled with some degree of virtualization – are necessary to address mid-size venues. At the same time, the ability to cover mid- size venues is crucial to the transition to massively densified networks, because of the amount of traffic generated by subscribers at these locations.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |14| 8. Network: Distributed / Centralized
One of the main decisions operators have to make as they densify their networks is how distributed or centralized the RAN and core functions should be to maximize spectrum and resource utilization, optimize performance, and keep costs down.
The dominant approach to RAN densification in today’s atomic networks – with the exception of DAS – starts with a phase in which densification is distributed: small and femto cells that include all eNB functionality – radio and baseband – are deployed at the edge of the network, often in places where the macro network has insufficient coverage or none at all. This approach makes small and femto cells fast and easy to deploy, because no central location is needed to host baseband functionality. Backhaul requirements can be easily managed because of the lower capacity and latency requirements of a centralized environment.
Centralized architectures, such as DAS, C-RAN and vRAN, present multiple advantages over distributed topologies. Cost benefits that accrue from concentrating baseband processing in a remote location apply to all RAN elements, from macro to small cells. Femto cells may benefit from some level of centralization, but typically the approach to virtualization is different because femto cells do not have cost- effective access to fronthaul. Cost is a key consideration in driving RAN centralization and virtualization of the macro infrastructure in the short term, but the performance and flexibility advantages are going to have a stronger impact in the mid to long term, especially in HetNet multi-layer environments. Cost savings are less for small- cell than for macro-cell deployments, because centralized topologies require fronthaul instead of backhaul to meet latency and capacity requirements, and fronthaul is typically expensive and accounts for a larger percentage of the capex and opex in a small-cell network than in a macro-cell network.
First among the benefits of a centralized architecture is the ability to manage transmission across multiple layers to minimize the effect of interference in co-
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |15| channel environments. Especially in outdoor environments where macro and small cells overlap in coverage and use the same spectrum channel, interference severely cuts into the capacity benefits of small cells; interference reduces the capacity not only of the small cells, but of the macro cells, which are more expensive and valuable on a per-bit basis.
Indoor small cells and DAS deployments suffer less from interference. The reduced coverage from the macro network, which is what drives in-building deployments, means interference is also more limited. New building codes that shield buildings from macro transmissions further reduce indoor coverage and interference, and indirectly foster a stronger commitment from mobile operators and the enterprise toward indoor deployments. In this environment, a centralized architecture is often beneficial, because it helps manage intra-layer interference, and it makes equipment installation and operation easier.
Centralized deployments also enable – but do not require – new ways to manage traffic in dense environments. Cell IDs can encompass multiple antennas and multiple RATs can be tightly integrated, such as in China Mobile’s no-more-cells model and in DOCOMO’s phantom cells. A centralized, virtualized RAN is better suited to load balancing traffic across antennas and wireless interfaces, and to managing network-sliced traffic, because all the processing is done in a single location where there is full visibility across the real-time load and availability of the locally available network resources.
Finally, a centralized architecture increases the efficiency of instruments like MEC that shift the core functionality to the edge – for instance, to lower latency (e.g., for video streaming), or to support venue-specific applications (e.g., local breakout for enterprise or IoT applications). In this case, deploying MEC in a C-RAN is less expensive and more efficient than in a distributed RAN, because less equipment is needed and better coordination can be achieved. For instance, content caching is more efficient in a centralized environment, where it is available to multiple access locations, than in a distributed RAN in which the caching is done at each access location.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |16| 9. Technology: Cellular / Wi-Fi
By far the most successful densification strategy to date has been Wi-Fi offload. It is Wi-Fi that has replaced wireline with wireless as the default access technology, not cellular. It was Wi-Fi that showed us what mobile broadband could do. When the iPhone came out, 3G networks did not have sufficient capacity to reliably support bandwidth-intensive services like video streaming, but Wi-Fi did, and it provided many subscribers the motivation to buy the new device.
According to Cisco VNI, Wi-Fi carries 43% of the IP traffic today, and VNI forecasts this percentage to grow to 50% by 2020. By comparison, cellular traffic accounts for 5% today, and that is forecast to be 16% in 2020. This means the dominant access technology for mobile devices is mostly outside operators’ control. Despite the growth in carrier Wi-Fi and, more generally, the increased push for Wi-Fi offload – for data, but also for voice with Wi-Fi Calling – it is the active choice of subscribers that has made Wi-Fi access prevalent, enabled by the wide availability of Wi-Fi infrastructure in residential and enterprise environments.
Wi-Fi is set to continue to play a crucial role as wireless networks densify, but at the same time the 2.4 GHz and 5 GHz spectrums it uses are getting congested – and the congestion will increase with the introduction of LTE access in the 5 GHz band (see below). Wi-Fi will expand to the 60 GHz band, but the dominant use cases there are for services in which the devices are in close proximity to the access point and fundamentally stationary.
As a result, Wi-Fi is likely to become the technology that, instead of providing offload, will need offload itself. An expansion in the allocation of unlicensed spectrum that Wi-Fi may use will bring relief, but we will also need greater availability, higher spectral efficiency and more intensive utilization of cellular bands. Densification can no longer be primarily entrusted to Wi-Fi; it will instead require the integration of multiple access technologies to provide the seamless connectivity that users expect.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |17| 10. Unlicensed spectrum: LTE / Wi-Fi
Wi-Fi’s success in winning the hearts of wireless users and in using spectrum with unprecedented efficiency, mostly because of dense deployments, has piqued the interest of vendors and operators that have seen an opportunity to use unlicensed spectrum for LTE.
The advantages to mobile operators are clear: operators gain opportunistic – i.e., not guaranteed, contingent on availability – access to unlicensed spectrum, using the same technology and network infrastructure they use for LTE, maintaining control over traffic and integrating it with their cellular RAN and core. In addition, LTE’s spectral efficiency is higher than Wi-Fi’s because of more efficient modulation. With the exception of MulteFire (see below), LTE in unlicensed bands is deployed alongside licensed LTE, and, as a result, the marginal cost of adding LTE unlicensed is low in greenfield small-cell deployments.
There are, however, disadvantages to using LTE instead of Wi-Fi in unlicensed bands as well. Wi-Fi is already installed in virtually all mobile-broadband devices. The infrastructure is widely available, and in most cases free to access – both to operators and to subscribers. In contrast, LTE in unlicensed spectrum requires new devices and new infrastructure that has to be deployed – and paid for, in most cases – by the operator. Besides, in order to use LTE in unlicensed bands without unduly affecting Wi-Fi performance, LTE has to use LBT mechanisms that reduce its performance advantages over Wi-Fi. Finally, deployments of LTE unlicensed require the consent of venue owners where they have control over the location of the LTE unlicensed antennas. Those owners might not grant permission to install, because LTE unlicensed – even when it uses LBT – competes with the local Wi-Fi for network resources, which venue owners consider theirs and want to continue to control.
It is still unclear how widely LAA, the version of LTE unlicensed that is designed to guarantee nice coexistence with Wi-Fi and which is deployable worldwide,
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |18| worldwiwill be adopted in the face of the competition from Wi-Fi and the LWA allows mobile operators to use unlicensed spectrum but operates in tradeoffs it requires. the opposite direction of LAA. It uses the Wi-Fi air interface, but it fully integrates the Wi-Fi traffic within the LTE network. This approach removes A further uncertainty is due to two promising – and to some extent the controversial coexistence of LTE and Wi-Fi in the unlicensed 5 GHz band, complementary – alternatives: MulteFire and LWA. but it also removes the LTE performance advantage.
MulteFire allows for the use of LTE as the air interface in the 5 GHz band, As long as devices can support LAA, MulteFire and LWA (and they likely will), without needing to use a licensed band for the control plane. So venue mobile operators will have multiple ways to use unlicensed spectrum in owners or operators that do not have an LTE network in licensed bands can addition to Wi-Fi – as long as they get access to the venues they want to deploy MulteFire and, if they choose, offer LTE unlicensed access on a cover. We expect operators to select one or more of these solutions, as neutral-host basis to mobile operators. This approach enables operators to soon as they deploy small cells, because using unlicensed spectrum not only improve capacity and/or coverage in venues where they may not have augments capacity, it significantly improves the business case for small cells, access or where they do not want to deploy licensed LTE. which to date has been a difficult one.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |19| 11. Spectrum: Sub-6 GHz / Millimeter wave
Massive densification requires new spectrum to manage the increase in traffic Millimeter-wave bands are the other hot prospect for spectrum expansion. The loads and meet performance requirements efficiently and cost effectively. By amount of spectrum available is huge and, because of the high reuse that high packing the infrastructure closer and using multiple layers (e.g., small cells and frequency makes possible, the potential increase in capacity is astounding. macro cells), operators can reuse a frequency channel more intensely. But at some Spectrum in these bands is going to be much less expensive, or even usable on an point they face a diminishing return: the marginal investment to increase capacity unlicensed or lightly licensed basis, and current 5G standardization efforts cover becomes too high for the increase in performance it brings. mmW bands both for fixed wireless links (e.g., backhaul or possibly fronthaul) and for access in areas of very high density (e.g., where even small cells or DAS are not Many operators face this situation in the macro RAN. They increase the density of sufficient, the extremely hot spots). MmW supports densification on two fronts – base stations, added MIMO and CA, and split sectors, and they reach a plateau. xhaul to small cells, and access from UEs – which can be combined when using in- The next step is to add small cells in the same band, but that introduces band xhaul. interference – especially if the small cells are located outdoors in areas covered by the macro layer. In some environments the interference can be managed Using mmW for access requires more than new antennas or distributed small cells. effectively, but it adds a cost in terms of effort and lost spectral efficiency. Because of the small coverage radius, mmW access can generate large numbers of handoffs, as users will cross the cell-edge border frequently and move from mmW Using multiple bands to cover the same location enables operators to lower the to sub-6 GHz cellular coverage from macro or small cells. Frequent handoffs create impact of interference and minimize costs. Low-frequency bands are still valuable high levels of overhead that affects the anchor sub-6 GHz network. for improving coverage in low-density areas and for some IoT applications. In high- density environments, higher-frequency bands are more effective in reducing the A solution to this problem is to implement a phantom-cell architecture, in which interference and increasing spectrum utilization, because of the more limited coexisting elements – e.g., sub-6 GHz and mmW – become part of the same cell ID coverage range. This is a major limitation in a macro network, and for this reason and the control plane is managed for both bands from the sub-6 GHz bands, which mobile operators have been reluctant to use spectrum above 2.5 GHz to date. have wider and more consistent coverage. Load balancing across bands enables operators to direct traffic to the band that can accommodate it more efficiently. Increasingly, however, mobile operators have become keen to use higher frequencies for lower layers. Candidates range from the 3.5 GHz band to the Further densification and more intensive spectrum utilization may come from unlicensed 5 GHz band, and all the way to mmW. The 3.5 GHz band is an excellent device-to-device communications – either subscribers or IoT devices. Device-to- candidate for small-cell deployments, because its shorter range strikes a good device communications can establish an ad hoc, mesh-like network that expands coverage/capacity balance. However, the amount of spectrum available is limited, the reach and capacity of the rest of the network. This possibility calls into and in some countries there are regulatory restrictions or spectrum allocation question the sharp boundary between network and device that is prevalent in issues that have been slowing down the plans to use the 3.5 GHz band. today’s networks.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |20| 12. Interference: Co-channel / Separate channel
Initial densification efforts with small cells and DAS have used a co-channel model result, operators have been reluctant to deploy small cells widely, and have in which they share spectrum with the macro layer, but generate different levels focused on areas where the need for additional capacity makes them less of interference. Femtocells, the first incarnation of small cells, were largely sensitive to cost. deployed indoors, at very low power and in places where macro coverage was weak or absent, and so interference was not a big issue. Two ways to cope with the interference issues have emerged and are likely to accelerate massive densification. The first of these involves spectrum bands that Outdoor small cells were initially deployed in many cases to address areas are not traditionally used for mobile and that, therefore, are less expensive and in without coverage. Interference was, by definition, not an issue. When small-cell higher-frequency bands – such as the 3.5 GHz, the unlicensed 5 GHz band and deployments moved to dense areas where operators had macro coverage but mmW bands. Operators can use these bands in addition to co-channel spectrum insufficient capacity, interference became a major issue that slowed the rollout of within the same small-cell enclosure, or they can deploy a sublayer in these bands small cells. It also pushed operators to spend more money on increasing macro and let the macro layer retain control of cellular spectrum. In both cases, the cost capacity, and to move to small-cell deployments as a last resort after all the per bit decreases. In the first case the low incremental cost of adding these bands, macro enhancements had been used. Initially, techniques like CoMP or eICIC coupled with the increase in capacity, reduces the per-bit capex and opex. In the were not sufficiently mature or widely deployed, making interference second case the business case is strengthened by preserving capacity in the management less efficient. But more fundamentally, interference between small macro layer by avoiding interference. and macro layers in co-channel deployments reduces the capacity of the macro layer, and that is highly undesirable and expensive for mobile operators. Even if A second way of coping with interference is to move the lower-layer the cost of installing a small cell is lower on a per-bit basis than a macro cell, the infrastructure indoors. In locations where per-small-cell deployment and cost savings could easily dissipate if the small-cell installation results in a operating costs are comparable or lower than in an outdoor environment, indoor reduction of the macro-layer capacity. networks can help reducing the per-bit capex and opex. The reduced impact from interference lowers the per-bit cost in indoor deployments. However, in-building Of course, an effective way to minimize interference is to move away from co- infrastructure faces business model, ownership, and control challenges that are channel deployments and use a separate channel. Operators largely resist this, different from outdoor infrastructure, and access to indoor locations may not be because typically cellular spectrum is too expensive to use only in the small-cell available or affordable to mobile operators. The move to in-building layer. In most environments, a co-channel deployment supports a more efficient infrastructure will undoubtedly intensify and accelerate the densification process, use of spectrum, in terms of capacity density (bits per sq km), than a deployment but it will not eliminate the need for outdoor infrastructure, which will still be in which small cells use a channel different from that used by the macro layer. But used to meet demand from outdoor subscribers and in places where operators the reduction of capacity from interference makes the business case for small cannot deploy indoor infrastructure. cells less palatable, because it increases the overall network per-bit cost. As a
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |21| 13. Backhaul: Fiber / Wireless
A vexed question in the small-cell ecosystem is whether backhaul should be wireline (and, if so, whether it has to be fiber) or could be wireless (and, if so, in which bands). The debate does not regard indoor deployments, where wireline backhaul is typically used within the building. In outdoor environments, however, both wireline and wireless solutions present benefits and limitations – and approaching backhaul as a combination of fiber and wireless links may help reduce the drawbacks of both technologies.
Fiber is the ideal backhaul for outdoor small cells, but it is not always available, and even when available, it is not always cost effective. In most environments, fiber is available in the vicinity of the small cells, but bringing it to the small cell is often too expensive, not to mention time consuming.
Wireless backhaul is the alternative, as long as small cells do not use a C-RAN or virtualized architecture. With remote baseband, fronthaul requirements are tighter and wireless fronthaul is possible, but it requires specialized solutions. Solutions that work for backhaul typically do not have enough capacity or a low enough latency to support fronthaul. The limitation of wireless backhaul is that to provide sufficient latency, line of sight from the small cell to the aggregation point is required. As the link length grows, the likelihood of having a reliable line of sight and reliable performance decreases. Multi-hop wireless backhaul can compensate for the lack of line of sight, but increases costs and latency.
In outdoor environments, fiber and wireless are not mutually exclusive, but rather two components defining the backhaul. In fact, a mix of fiber and wireless backhaul is the dominant solution, with the choice between the two dictated by cost and availability tradeoffs. The backhaul for every small cell terminates into fiber; what changes is the length of the wireless backhaul, which can be zero for a small cell with fiber backhaul.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |22| 14. Fronthaul: CPRI / Xhaul
Another concept that pervasive networks challenge is the dichotomy of backhaul and fronthaul in terms of requirements and what technologies could meet them, and in terms of what type of signal the fronthaul carries.
Within mobile networks, fronthaul is the link from the RRH to the BBU, and backhaul the link from the BBU to the core network. If the RRH and BBU are co-located, there is no need for fronthaul. Because the fronthaul carries the analog signal, the bandwidth and latency requirements are much higher than those for backhaul. As a result, some technologies and solutions may be suitable only for backhaul. Others may be well-suited for fronthaul, but too expensive for backhaul.
First off, with the move to extra-low latency and high capacity (e.g., when using mmW for access), the 5G backhaul requirements may approach those for fronthaul today. But this also means that fronthaul requirements, if using CPRI, will also grow to alarming levels. This creates the possibility that the fronthaul may become the bottleneck, and the risk that RAN capacity may have to be capped if the BBUs are remote. (Another possibility is that RAN virtualization will slow down because fronthaul requirements are too onerous.)
Because it is difficult to meet current and future fronthaul requirements with CPRI, especially in small-cell deployments where dark fiber is too expensive or not available, there is considerable ongoing work to define alternative interfaces – e.g., compressed CPRI or Ethernet – and functional splits, in which some of the baseband functionality stays at the edge, co-located with the RRH.
With functional splits, the fronthaul requirements decrease, but so do the benefits of virtualization. Multiple options are available, and operators need to decide what level of functional split works best in each location in their networks. It is not yet
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |23| clear whether there will be a dominating functional split or which splits will be best suited to which environments.
Importantly, functional splits call into question a clear-cut distinction between fronthaul and backhaul, and instead suggest a continuum of functional splits between RRH and BBU, which recently has been referred to as the xhaul. Then, depending on the type of xhaul – i.e., the selected functional split – and the RAN requirements, different interfaces and solutions become appropriate. The xhaul approach recognizes operators’ need for flexibility as they densify their infrastructure and use a wide range of RAN solutions to strike the right balance between costs and performance.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |24| 15. Access point density: High / Low
When planning for densification, operators need to select not only technologies, solutions, spectrum bands and locations, but also the topology of the densification. There, access point density is a choice that is becoming more prominent, although it is rarely discussed as a stand-alone dimension of the densification strategy. As a wider range of solutions, spectrum bands and business models become available, operators can choose among options that vary in power, coverage, cost, equipment size, and ability to fit into multi-operator, neutral-host models.
Given a target capacity they aim to have in a given location, operators can choose to have a high number of low-power, reduced-coverage, low-capacity access points, or a smaller number of high-capacity access points. When using mmW for access, or multiple bands in the same access point, operators can create super hotspots where they have a very high concentration of traffic.
The primary consideration in selecting the appropriate access-point density is the distribution of traffic. Subscribers and the traffic they generate are distributed very unevenly, so it is crucial to place the access points as close as possible to subscribers. If the distribution is uneven and highly clustered, access points will be placed more densely in high-usage locations. If the distribution is more even, access points with larger coverage areas may be preferred.
In addition to traffic distribution across locations, operators need to consider backhaul availability, deployment costs, spectrum availability (including traffic load in unlicensed bands), business model (e.g., whether there is infrastructure sharing among operators or a neutral-host model) and RF propagation in the environment. For instance, if deploying access points is inexpensive, backhaul is easily available, or spectrum availability is limited, an operator may opt for a denser network of access points with limited coverage and capacity. Fewer but more powerful access points may be better suited in locations where installation and backhaul are expensive.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |25| 16. Business model: Single operator / Shared deployment
Densification demands that the wireless infrastructure move closer to revenues or to get better mobile service or both, but they do not know how to subscribers, vehicles, and IoT sensors and other devices – and this means structure deals with operators because it is uncharted territory. The same is beyond the cell-tower model, which continues to be used for the macro layer true for mobile operators. As a result, we see trials, negotiations and even legal but needs new indoor and outdoor models as complements. challenges – but little in the way of the expected large deployments. The urgency of densification is increasingly felt by both parties as bad wireless In a macro environment, the operator has full control over network planning, service affects operators and venue owners or administrators, and so deals are deployment and operations, using telco assets such as cell towers. The starting to come together. They will undoubtedly evolve through time as we operator owns the end-to-end infrastructure and manages traffic and understand the deployment models better from technical and business interference. Cell towers are usually owned by third parties, but operators lease perspectives, but these early deals are the necessary first step to get beyond ad space on them and retain control of the telecom equipment. This model, well hoc densification to large-scale densification. understood by every player in the ecosystem, ensures that deployments proceed smoothly. More interestingly, the need for mobile operators and asset owners to work together – directly, or indirectly through third parties such as service providers, In a sublayer deployment, this model no longer works. The access fiber providers or cell-tower companies – is also accelerating the development infrastructure has to be mounted on non-telecom assets, such as lampposts, of closer relationships with cities, enterprises and other venue owners and exterior or interior building walls, ceilings, advertisement displays, and public institutions. These relationships can lead to better performance and to the transportation vehicles, or even below street level. This creates constraints on creation of services that mobile operators can develop or support. In a where operators can deploy the equipment; it also creates the need to densified, pervasive network, it is not just the equipment that moves closer to establish direct or indirect relationships with the owners of these assets – cities subscribers – it is the relationship among subscribers, venue and public entities, enterprises, educational institutions and other venue owners/administrators, and operators that becomes tighter and deeper. owners. In some cases, these entities expect to extract rent from mobile operators or third parties working on their behalf. In other cases, the asset The need to negotiate deals with asset owners coupled with the need to find owner may pay for the infrastructure or encourage rent-free installation and cost-effective ways to deploy small cells drives new business (and deployment) operation, but may require visibility or some level of control over the local models. The initial small-cell business model assumed each operator would network. deploy its equipment independently of other operators or service providers, striking deals with asset owners. But that is too expensive, time consuming and Negotiating deals with the new asset owners has proven to be one of the inefficient to scale beyond prime locations (e.g., some parts of Manhattan, biggest challenges – and causes of delay in deployments – that operators face. downtown San Francisco, central London) to areas with less concentrated, but Asset owners are eager to host the telecom infrastructure, either to extract still high, traffic.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |26| Especially in indoor venues, the trend is toward shared deployments, in which This model does not exclude opportunities for the operators and venue owners one operator or, more commonly, a neutral host acts as the interface between to negotiate specific arrangements about network performance, coverage or the asset owner and the operators that wish to participate in the local network. functionality, or to provide additional services (e.g., enterprise services, or IoT services to city agencies). These arrangements will encourage the venue This is very similar to the DAS neutral-host model, in which multiple operators owners to participate more actively in the densification process – in part by can share the DAS, but each controls its own transmissions. This model enables funding deployments, and in part by requesting services – and to see the operators to reduce costs by sharing some of the infrastructure, and it wireless infrastructure as an integral component of the venue, comparable to streamlines deployment and operations, because the neutral host manages all the in-building electrical network. the relationships with venue owners on one end, and with multiple operators on the other end. There is a widespread belief that the DAS neutral-host model In this context, the availability of the 3.5 GHz band in the US and of other sub-6 does not work in small-cell deployments because it requires operators to share GHz bands in other countries will push the shared deployment business model the RAN – an option that nearly all operators consider unacceptable if it further, because it allows neutral-host players to install a small-cell network in a involves their licensed spectrum. But the neutral-host model can equally allow venue, using spectrum bands that are not allocated to a specific operator. The operators to deploy their own radios, retain control of the use of the spectrum venue owner or the neutral host can deploy the network and allow access to and manage traffic, without having to negotiate a separate deal with the asset operators on a wholesale basis. A benefit of this approach is that the neutral owners, and without having to directly deploy and manage the RAN host or venue owner can deploy a single network that can be shared, and this equipment. results in more-efficient spectrum utilization and in lower costs.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |27| 17. Concluding thoughts. The role of operators in a densified, pervasive network
Many mobile operators worldwide question their future relevance – and worry about their ability to extract the revenues they deserve from their networks. Typically, these concerns are rooted in the risk they perceive that they will become a dumb pipe that can transmit an increasing amount of data to their subscribers and do so more reliably than in the past, only to see subscribers value the networks less than the OTT apps they use on those networks.
The transition to massively densified, pervasive networks can change this. The role of the mobile operator is transformed by the increased complexity of wireless networks that have to optimize and integrate multiple spectrum bands, technologies, services, device types and topologies. Mobile operators can no longer focus only on pushing as many bit/s as their infrastructure supports; they also need to manage traffic, applications and network resources in a much smarter way than they are accustomed to. Increasingly, they do not look like a utility – they look instead more like orchestra conductors, coordinating transmission in a multi-layer network – a network in which they have the flexibility to set strategy in ways that differentiate their network from that of their competitors.
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |28|
II. Vendor profiles and interviews
REPORT Massively densified networks © 2016 Senza Fili www.senzafiliconsulting.com |29| Tools helps mobile operators with documentation Anritsu solutions also help operators with Profile and reports, real-time analytics, automated densified networks to identify the different assessment of RF sweeps and PIM test results. sources of interference that affect macro-cell and Anritsu small-cell networks and to manage interference, if The ability to automate and scale testing and necessary, in real time. monitoring in wireless networks is crucial for Founded over 120 years ago, Anritsu Corporation operators moving to multi-layer, multi-RAT The portfolio of Anritsu measuring instruments is is a global provider of communications test and networks and with DAS; the number of tests well suited for indoor densified networks. It measurement solutions. needed to assess performance rapidly increases consists of solutions for both the wireless and with complexity and makes manual field testing optical segments; they can test and monitor both Anritsu’s measuring instruments support multiple time consuming and expensive in terms of staff the access and backhaul/fronthaul portions of areas: resources. mobile networks.