1

Ultra-Dense Networks in 5G A Survey Paper

Gaurav Verma(14244), Maniratnam Mandal(14365), Siddhant Rohela(14690)

Department of Electrical Engineering Indian Institute of Technology, Kanpur

Abstract—The extreme traffic load that future wireless

networks are expected to accommodate requires a re-thinking of the system design. Initial estimations indicate that, different from the evolutionary path of previous cellular generations that was based on spectral efficiency improvements, the most substantial amount of future system performance gains will be obtained by means of network infrastructure densification. By increasing the density of operator-deployed infrastructure elements, along with incorporation of user-deployed access nodes and mobile user devices acting as “infrastructure prosumers”, it is expected that having one or more access nodes exclusively dedicated to each user will become feasible, introducing the ultra-dense network (UDN) paradigm. Although it is clear that UDNs are able to take advantage of the significant benefits provided by proximal transmissions and increased spatial reuse of system resources, at the same time, large node density and irregular deployment introduce new challenges, mainly due to the interference environment characteristics that are vastly different from previous cellular deployments. This article attempts to provide insights on fundamental issues related to UDN deployment, such as determining the infrastructure density required to support given traffic load requirements and the benefits of network-wise coordination, demonstrating the potential of UDNs for 5G wireless networks.

I. INTRODUCTION

Over the next decade, the wireless traffic volume is expected to increase thousand times over. To meet the demands, the fifth generation (5G) cellular network is becoming a hot research topic in telecommunication industries and academics.

Firstly, the massive multiple-input multiple-output (MIMO) technology was proposed to improve the spectrum efficiency of mobile communication systems [1]. Secondly, the millimeter wave communications was presented to extend the transmission bandwidth for 5G mobile communication systems [2]. Furthermore, the small cell concept has been appeared to raise the throughput and save the energy consumption in cellular scenarios [3]. It is being planned that to ensure seamless coverage, a larger number of small cells will be densely deployed for 5G cellular networks. Since the study of ultra-dense cellular networks is still in an initial stage, it becomes important to to investigate the network architecture and cellular densification limits for future 5G cellular networks.

In the 3G cellular networks, the densification of macro-cell base stations (BSs) aims to improve the transmission rate in partial areas, and the density of deployed macro-cell BSs is about 4-5 BSs/km2. In the 4g cellular networks, where microcell BSs have been deployed to satisfy the high speed transmission in specified regions, the density of microcell BSs is

approximately 8-10 BSs/km2. In 3G and 4G cellular networks, the aim of BSs densification is to improve the wireless transmission rate in partial regions. In 5G cellular networks, the massive MIMO antennas will be integrated into BSs, where hundreds pf antennas are utilized for transmitting Gbits level wireless traffic.

While the 5G BS transmission power is constrained at the same level of 4G BS transmission power, every antenna transmission power at 5G BS has to be decreased 10-20 times compared with every antenna transmission power at 4G BS. To satisfy the seamless coverage, the density of 5G BS is highly anticipated to come up to 40-50 BS/km2. Therefore, the future 5G cellular network is an ultra-dense cellular network.

II. MOTIVATION

With the arrival of every new generation of wireless communication, there has been a major overhaul in the technology used to provide blazing fast speeds to an ever increasing number of users around the planet. Up till the 4th Generation (LTE) of wireless technology, the focus of research has been on developing new and more robust components, as well as on spectral efficiency and bandwidth improvements. And this rigorous research has indeed borne astonishing results, with every new generation being many times faster than its predecessor. But the scenario is changing very rapidly. Our aim to digitally connect virtually everything around the globe, coupled with the fact that the volume of wireless traffic will increase by at least a 1000 times in the next decade; has compelled us to think along a different path. The attention has shifted to the densification of network infrastructure. The idea of densifying the infrastructure is not a new one. In fact, throughout the development of previous generations, this technique was employed as a ‘bonus’ to neutralize overloading at certain locations, hence rendering it a very regional concept.

2

But now the time has arrived to bring the UDN technology to the fore, in order to be able to meet the fantastic challenges that 5G has to offer.

Another strong motivation to employ UDN is its ability to utilize User Equipment (UEs) as not just consumers, but as a more comprehensive set of ‘infrastructure prosumers’ (producers + consumers). This is achieved by exploiting the advanced capabilities of modern day mobile devices, and handing them the role of infrastructure Access Nodes (ANs), along with the traditional ones. This will dramatically increase the number of active ANs, automatically amplifying the density of the existing infrastructure. It is predicted that the number of such ANs may even surpass the number of UEs in the coming years.

III. IMPLEMENTATION

In theory, the UDN 5G Technology is capable of catering to the ever increasing demands of the future generation of internet users. However, a robust architecture is required to implement this technology into actual use. For developing the UDN network, we take inspiration from the Conventional Cellular Network Architecture that is currently used for the LTE Technology. In the conventional architecture, a tree network of macro-cell Base-Station is embedded with a network (tree type) of microcells, both sharing management data as well as the user data traffic. Thus, the user has the freedom to choose between microcell BS and macro-cell BS, as per their requirement. Logically, the microcell BS is used in indoor scenarios, e.g. hotspots.

Fig: The network is densified by deploying small cells indoors in buildings and stores, and outdoors on trees, lampposts, and building walls. Small cell networks coexist with macro-cells, either in the same spectrum or on a dedicated carrier.

In the UDN architecture, it is extremely difficult to transmit traffic via fiber links (which was the case with the conventional network), partly due to the complex geography of urban areas (where 5G is primarily expected to function) and partly due to the high cost of installation. Instead, we use the millimeter wave technology to transmit traffic wirelessly. However, the small cell BSs using this technology have an effective range of transmission of only about 100 meters or so. Thus, the need arises to install a dense network of small cell Base-Stations, so that the data can be effectively relayed without much attenuation. Unlike the Conventional Network, the macro-cell BS and the microcell BS do not overlap in terms of data transmission. While the former takes care of transmitting the

management data, the latter is configured to solely transmit user data. Depending on the number of gateways in the macro-cell BS, there exist 2 types of network architectures:

A. UDNs with Single Gateway

In this type of architecture, only one gateway is installed at the macro-cell BS. Along with it, large antennas (MIMO millimeter wave) are also erected to receive backhaul traffic from the microcell BSs. Whole of this backhaul traffic is the ready to be forwarded to the core network. This time the transmission occurs by Fiber to cell (FTTC) links.

B. UDNs with Multiple Gateways

Apart from the multiple ones at macro-cell BS, multiple gateways are installed at several microcell BSs, which are judiciously chosen based on their relative geographies. Here, the traffic of small cells is transmitted to the multiple gateways in the macro-cell. All the backhaul traffic from the chosen microcell BSs and the macro-cell BS is collected and sent to the core network via FTTC links.

Fig: Single Gateway Architecture

Fig: Multiple Gateways Architecture

The small cells in UDN can be classified into fully-functioning base stations (BSs) (pico-cells and femtocells) and macro- extension access points (relays and Remote Radio Heads (RRHs)). The fully-functioning BS is capable of performing all the functions of a macro-cell with a lower power in a smaller coverage area. Specifically, the fully-function BS performs all the functions of the entire protocol stack. The table in the next page summarizes the features of different small cell types. In what follows we explain the different types of small cells. Horizontal and Vertical Densification The densification of wireless networks takes place mainly by the deployment of increasing number of small cells.

3

Accordingly, the network densification comes in two different flavors, either horizontal densification or vertical densification. In horizontal densification, the access nodes are densified in the horizontal plane, e.g., in the streets or hotspots. On the other hand the vertical densification evolves in the elevation plane where the customers deploy BSs in their apartments, offices, meeting rooms, and interior of buildings. The figure depicts vertical and horizontal network densification. The aforementioned classification highlights three major aspects of the densification schemes, specifically, the modelling of the densification scheme (vertical/horizontal) to consider appropriate propagation models, the backhauling alternatives for the different densification schemes, and the performance evaluation of the corresponding densification scheme.

IV. COMPARISON WITH CONVENTIONAL CELLULAR NETWORKS

The detailed differences between conventional cellular networks and 5G ultra-dense cellular networks with single/multiple gateways are explained as follows: The architecture of conventional cellular networks is a centralized network architecture and some microcells are densely deployed at partial areas, for satisfying crowded people communication requirements. When 5G small cell BSs equipped with massive MIMO antennas and millimeter wave communication technologies, the coverage of small cell has to be obviously reduced. To realize the seamless coverage, 5G cellular networks must be densely deployed by a large number of small cells. In this case, 5G ultra-dense cellular networks can provide the high bit-rate in all cellular coverage regions. Moreover, the architecture of ultra-dense cellular networks is a distributed network architecture considering cost and geography deployment requirements. Every BS in conventional cellular networks has the same function and the coverage between macro-cells and microcells is overlapped. For 5G ultra-dense cellular networks, macro-cell BSs transmit the management data and small cell BSs take charge of the user data transmission. There does not exist the overlap for the function and the coverage between macro-cell BSs and small cell BSs. Besides, 5G ultra-dense

cellular networks with single gateway are cost efficient but the backhaul capacity bottleneck maybe exists at the single gateway. 5G ultra-dense cellular networks with multiply gateways have to spend the high cost in the small cell deployment. Compared with conventional cellular networks, 5G ultra-dense cellular networks performance will provide graceful degradation as the degree of mobility increases. To overcome this challenge, the multi-cell cooperative communication is a potential solution for 5G ultra-dense cellular networks. The comparison between the conventional cellular networks and between Ultra Dense Networks implemented using single gateway and multiple gateways has been summarized in the following table:

V. ADVANTAGES OF ULTRA-DENSE NETWORKS

A data analysis of the Flickr social online network (Scholz 2015).

Network Types Conventional Networks

UDNs with Single Gateway

UDNs with Multiple

Gateways

Network architecture

Centralized architecture

Distributed architecture

Distributed architecture

Densification deployment

target

Macro-cells Small cells Small cells

Functions of macro-cell and

microcell

same Macro-cells transmit

management data, microcells

transmit user data

Macro-cells transmit

management data, microcells

transmit user data

Backhaul method

Backhaul traffic is directly

forwarded into the core network

by gateway

Backhaul traffic is relayed to the

gateway by multi-hop wireless links

Backhaul traffic is relayed to the

gateway by multi-hop wireless link

Number of backhaul

gateway in a macro-cell

One One Multiple

Merit Flexible deployment and

low cost

Ubiquitous and high bit-rate

Ubiquitous and high bit-rate

Demerit Small cell partial deployment, low network capacity,

uneven distribution of the achievable

data

Low mobility and there exists the

backhaul capacity bottle-neck

Low mobility and high cost

4

This study shows that the network link structure changes with increasing connectivity. Sparsely connected networks (left) show the typical power-law (scale-free) node-degree distribution in which most nodes have only few links while some few nodes are extremely linked. By contrast, densely connected networks (right), show a more “fair” distribution of less differences between lowly and highly linked nodes.

Ultra-Dense Networks are able to take advantage of the significant benefits provided by proximal transmissions and increased spatial reuse of system resources. Their purpose is to provide more capacity through offloading where needed. With the advent of ultra-dense networks combined with the diminished role of the legacy macro cell, handsets in close vicinity of each other can communicate directly through device-to-device links. UDNs bring with them a huge number of advantages over existing technologies which extend from minimized cost, extremely high speeds and operational efficiency to group broadcast and mesh networking, to enhanced security provided from the absence of intermediate network routing.

A. Minimized Cost and Higher Speeds:

Macro sites provide more capacity through multicarrier, carrier aggregation and advanced antenna solutions using available transmit power. Existing sites can be upgraded to support more capacity at lower cost than establishing new sites.

Small cells to provide lower cost capacity and coverage at outdoor hot spots.

Small cells to offload the macro network and provide low cost indoor capacity.

Indoor DAS solutions can be upgraded to LTE for seamless connectivity.

A study conducted by Nokia in Tokyo concluded that network capacity in a dense city can be increased up to 5-700x compared to 2010 using only outdoor deployments. Basic coverage is provided by the existing macro cells, street level capacity is provided with outdoor small cells and high rise capacity with up-tilted macro cells from existing sites.

Indoor deployments increases capacity even further, most likely through evolution of both indoor small cell and outdoor macro and small cell deployments. In its case study, Nokia showed that 1000x capacity can be achieved using both outdoor

small cells and indoor small cells. 100 outdoor small cells require 1,700 indoor small cells, while increasing the outdoor small cells to 200 cuts indoor cells to 200. This allows a tradeoff between indoor and outdoor deployments depending on local factors. Wi-Fi was deployed indoors but LTE small cells would perform similarly. The Tokyo deployment case corresponds to the definition of UDN for 2020 and beyond. Figure shows an example of the capacity of different indoor solutions in the building, with each scenario providing 95 percent coverage. The first case uses DAS for indoor coverage and the second uses the DAS infrastructure with a small cell on every floor, doubling capacity. Deploying further small cells improves capacity significantly. The final case combines DAS in the common area and small cells in the offices.

The indoor deployment with distributed small cells provides significantly more capacity, with the same coverage as DAS, at lower cost.

B. Operational Impact:

An operator in a medium sized country may have 8-10,000 base station sites, but as the network becomes denser, the operator may have more than 100,000 cells to manage in a complex environment. With this increase comes an exponential rise in network data. Yet the operator will need to keep operational expenses low using the same number of engineering staff. Solving this will involve analytics, extensive automation and redesigning tasks and processes.

Failure of a macro site can affect many subscribers and requires urgent attention of an expensive field engineer. However, with small cells, the same type of fault can have a much smaller impact on service availability because the problem is localized, and the area may be served by other cells including the macro umbrella. Operators that have adopted a Service Operations Center (SOC) can use advanced service quality management tools and processes to quickly establish the precise effects of the fault. Instead of responding immediately to individual faults, it may be possible to resolve these as part of a routine maintenance call.

C. Group Broadcast and Mesh Networking:

In outdoor wireless networking, wireless mesh networks are the third topology after point-to-point and point-to-multipoint in order to build a wireless network infrastructure. The major advantage of a wireless mesh networks is the intrinsic

5

redundancy and, consequently, reliability because a mesh network is able to reroute traffic through multiple paths to cope with link failures, interference, power failures or network device failures. Each node in the network is referred to either as Mesh Router (MR), Mesh Client (MC), or Mesh Gateway (MG) depending on its role in the network. MRs are interconnected to form a mesh backbone network, which can relay communications service from MCs to the MG. The advantages are given as:

A broken node won’t distract the transmission of data in a mesh network. Each node is connected to several other nodes which make it easier to relay data. A broken device will be ignored by the signals and will then find a new one that is connected with the node.

Additional devices in a mesh topology will not affect its network connection. Hence it will improve the traffic in the network. Mesh topology makes a large data center that simulates useful information to its nodes.

Can handle high amount of network traffic since every additional device into the network is considered a node. Interconnected devices can simultaneously transfer data smoothly and will not complicate the network connection.

Because all nodes in a mesh are receiving and translating information, there is some redundancy in a mesh topology; however, you can also gain speed with the excess bandwidth. If one route happens to be slow, a mesh network could potentially find a better route and optimize itself.

VI. SHORTCOMINGS AND CHALLENGES

As discussed in the above sections, massive MIMO antenna and millimeter wave communication technologies paved the way for ultra-dense cellular networks. Ultra-densification of network is considered the main enabler for 5G wireless access. The future envisioned scenarios involve average rates of more than 100 Mbps/user in order to support video content, possibly at an 8K resolution, or accessing multiplayer cloud gaming systems. The UDN paradigm aspires to revolutionize current cellular thinking by efficiently exploiting the advantages offered by proximal communications and sets an ambitious goal of increasing the capacity per area of current 4G network by 1000 fold. The main challenges that come with dense cell deployment include interference, mobility, power consumption and backhaul. In the ideal case, capacity should scale linearly with TP density but issues including interference and mobility may severely hamper the gain of densification. Moreover, practical deployment and maintenance related aspects, such as backhaul and energy provision, are also crucial for the feasibility of UDN.

The first challenge is the multi-hop relay optimization in 5G ultra-dense cellular networks. In the distribution network architecture, both the back-haul and front-haul traffic must be relayed into the destination. The selection of relaying small cell BS should be carefully considered in 5G ultra-dense cellular networks. Hence, the wireless multi-hop routing algorithm is a key challenge for 5G ultra-dense cellular networks. Although the small cell BS equipped with massive MIMO antennas has

enough antennas for simultaneously transmitting backhaul traffic and front-haul traffic, it is another important challenge how to reasonably allocate massive antennas for backhaul and front-haul transmissions.

The second challenge is to figure out how to organize adjacent small cells for co-operative transmission. For example, how to dynamically group small cells for seamlessly covering the high-speed mobile user track is an open issue. The proportion between the computation power and transmission power maybe reversed at wireless transceivers adopting massive MIMO antenna and millimeter wave communication technologies. In this case, the computation power cannot be ignored for the BS energy consumption. Considering the proportion change between the computation power and the transmission power, the new energy efficiency model need to be investigated for ultra-dense cellular networks with massive MIMO antenna and millimeter wave communication technologies.

Thirdly, as a perennial problem faced by every generation of mobile communication system, interference is serious in LTE since adjacent cells reuse the same frequency resource. To cope with the interference between macro cells, some static and semi-static methods can be proposed as of now. Static methods such as cell planning and partial frequency reuse (PFR) use system configurations to avoid interference. In semi-static methods, eNBs exchange information for interference coordination via X2 interface with a period ranging from 20ms to 200ms. For downlink, relative narrow band transmit power (RNTP) is utilized to inform neighboring eNBs of the bands with high power. For uplink, overload indicator (OI) signals measured interference level while high interference indicator (HII) reports the bands on which high power uplink transmission may be scheduled.

Fourthly, Mobility is another important issue which is usually handled by handover in mobile communication systems. Handover is not expected to happen frequently as they were initially designed for macro cells with large coverage in LTE. Thus a complicated handover procedure with high signaling overhead is adopted. Moreover, the hard handover mechanism in LTE means connection to the source cell is released before connecting to the destination cell. This disruption in transmission may trigger the congestion control process of transmission control protocol (TCP), leading to violent fluctuation of data rate.

Lastly, energy consumption delivers a difficult challenge when it comes to UDNs. Statistics show that mobile communication industry produces 86 million tons of CO2 in 2007, accounting for 0.2% of global carbon footprint. Even with an optimistic estimation, mobile communication carbon emission is expected to triple by 2020. Since UDN means much more small cells, the surge in energy consumption will be considerable if the 4G energy efficiency is not improved. High energy consumption not only leads to environmental damage but also puts pressure on the already huge bills of the operators. Moreover with the advent of green energy, green communication is becoming a hot research topic and expectation for 5G to be more energy efficient than previous technologies is becoming a significant drive.

6

VII. ENABLING TECHNOLOGIES AND NEW TRENDS

Facing the aforementioned challenges, both academia and industry devote a lot of resource to develop effective solutions. In this section, the technologies we view as the most promising components of 5G UDN are discussed. The important fact is that one technology may help in solving one or more challenges to some extent.

1. Cell and receiver virtualization: Virtualization refers to network function virtualization (NFV), which virtualizes logical network entities on general purpose hardware. While NFV is mainly implemented in core network, similar concept may also be applied to radio access network and the result is cell and user virtualization. As illustrated in the Fig, several TPs around a user can form a virtual cell (VC) to provide service jointly. The TPs transmitting to and receiving signal from UE are called serving TPs. One master TP (MTP) is selected to control the operations of a VC if centralized control is adopted. It needs to be noted that MTP is not necessarily a serving TP and a TP can join multiple VCs simultaneously. MTPs of adjacent VCs can negotiate on matters such as resource allocation to enable the harmonious coexistence of different VCs. Since cooperative TPs are in proximity, fast coordination and negotiation are possible with self-backhaul or device to device communication (D2D). VC can help in various aspects using fast coordination, such as interference management, mobility management and energy conservation.

Illustration of cell virtualization

User virtualization with carrier sharing

2. Self-Backhaul: Self-backhaul has the advantage of low cost, controllable link quality and inherent support for semi-planned or unplanned deployment. With the potential of cutting the delay of inter-cell communication down to one sub-frame, self-backhaul can also facilitate delay sensitive inter-cell coordination. Although a

mixture of various backhaul technologies is expected for 5G UDN, we still believe self-backhaul will play an important role. However, substantial work needs to be done to improve Release10 relay in terms of capacity and flexibility. Several designs can be used to enhance the capacity of self-backhaul. In most UDN scenarios, radio link is LoS due to short link distance or high carrier frequency. Without sufficient multipath components, it is hard to transmit multiple data flows

simultaneously to a receiver even when multiple antennas are equipped. A strategy named BC (broadcast channel)+MAC (multiple access channel) is designed for this scenario and plotted in the figure.

Illustration and performances gain of BC+MAC

Network coding can also enhance the throughput of self-backhaul and an example is provided in figure below. The network coding scheme is conducted in two steps. In the first step, both UE and dTP transmit their signal to rTP. Normally rTP would forward the message of UE to dTP and vice versa. Considering the broadcast nature of wireless channel and the fact that both UE and dTP know their own message, rTP can simply transmit a XOR version of the two messages. Utilizing the local message and received message, the intended message can be reconstructed.

Network coding for capacity enhancement in self-backhaul

7

3. Use-centric access: In previous generations of mobile communication system, system information and common signal are broadcast to provide UE with access information. Moreover, CRS broadcasted every sub-frame can take up 14% of the overall OFDM symbols when four antenna ports are used. In UDN, most cells will be light loaded for a large portion of time. Periodic broadcasting of these common signals not only wastes energy but also causes strong inference to adjacent cells. In user-centric design, periodic transmitted signal is minimized, which means only the information crucial for initial access is broadcasted. Information such as cell reselection and MBSFN setting is provided in an on demand manner. UE can either require them in the random access process or using uplink control channel when connection is established. Another option is to transmit system information in a hierarchical manner. That is, the information common to a number of cells and vital for initial access can be broadcasted in single frequency network (SFN) manner by several TPs to improve coverage and save energy. Cell-specific information is broadcasted by each cell with the right to override information broadcasted using SFN. Traffic related and rarely used system information can be provided only when required.

REFERENCES

[1] “Ultra Dense Network: Challenges, Enabling Technologies and New Trends”: Hao Peng, Yan Xiao, Yu-Ngok Ruyue, Yuan Yifei, ZTE Corporation Shenzhen, China, 2016

[2] “Ultra Dense Networks: The New Wireless Frontier for Enabling 5G Access”: Antonis G. Gotsis, Stelios Stefanatos, and Angeliki Alexiou, 2015

[3] “5G Ultra-Dense Cellular Networks”: Xiaohu Ge, Senior Member, IEEE, Song Tu, Guoqiang Mao, Senior Member, IEEE, Cheng-Xiang Wang, Senior Member, IEEE, Tao Han, Member, IEEE, 2015

[4] “Supporting mobility in 5G: A comparison between massive MIMO and continuous ultra dense networks”:2016 IEEE International Conference on Communications (ICC), Written by: Petteri Kela; Xavier Gelabert; Jussi Turkka; Mário Costa; Kari Heiska; Kari Leppänen; Christer Qvarfordt

[5] “User-Centric Ultra-Dense Networks for 5G: Challenges, Methodologies, and Directions”: Shanzhi Chen, Fei Qin, Bo Hu, Xi Li, and Zhonglin Chen, 2016

[6] “Ultra Dense Networks” White Paper- Nokia [7] “Ultra-Dense Networks: A Survey”: Mahmoud Kamel, Student Member,

IEEE, Walaa Hamouda, Senior Member, IEEE, and Amr Youssef, Senior Member, IEEE, 2016

[8] “Getting connected - The highly connected society”: Network-Science.org

[9] “Wireless Mesh Networks”: fluidmesh.com [10] “Mesh Router Selection to Maximize System Throughput in Dense

Wireless Mesh Networks”: Panu Avakul, Hiroki Nishiyama, Nei Kato, Yoshitaka Shimizu, Tomoaki Kumagai, 2013

[11] “Advantages and Disadvantages of Using Mesh Topology”: networking-basics.net