Cloud radio access networks (C-RANs) provide a promising architecture for the future mobile networks needed to sustain the exponential growth of the data rate. In C-RAN, one data processing center or baseband unit communicates with users through distributed remote radio heads, which are connected to the baseband unit (BBU) via high-capacity low-latency so-called fronthaul links. The architecture of C-RAN, however, imposes a burden of fronthaul bandwidth because raw I/Q samples are exchanged between the RRHs and the BBU. Therefore, signal compression is required on fronthaul links owing to their limited capacity. This chapter exploits the advance of joint signal processing to reduce the transmission rate on fronthaul uplinks. In particular, we first propose a joint decompression and detection (JDD) algorithm which exploits the correlation among RRHs and jointly performs decompressing and detecting. The JDD algorithm takes into consideration both fading and quantization effects in a single decoding step. Second, the block error rate of the JDD algorithm is analyzed in a closed form by using pairwise error probability analysis under both deterministic and Rayleigh fading channel models. Third, on the basis of the analyzed block error rate (BLER), we introduce adaptive compression schemes subject to quality of service constraints to minimize the fronthaul transmission rate while satisfying the predefined target QoS. The premise of the proposed compression methods originates from practical scenarios, where most applications tolerate a non-zero BLER. As a dual problem, we also develop a scheme to minimize the signal distortion subject to the fronthaul rate constraint. We finally consider the counterparts of these two adaptive compression schemes for Rayleigh-fading channels and analyze their asymptotic behavior as the constraints approach extremes.

Introduction

Cloud radio access networks have been widely accepted as a new architecture for future mobile networks to sustain the ever increasing demand in the data rate [1]. In a C-RAN, one centralized processor or BBU communicates with users distributed in a graphical area via a number of remote radio heads (RRHs), which act as “soft” relaying nodes and are connected to the BBU via high-capacity and low-latency fronthaul links. By moving all baseband processing functions from RRHs to a centralized processor, the C-RAN enables adaptive load balancing via a virtual base station pool [2] and effective network-wide inter-cell interference management thanks to multi-cell processing [3, 4].

Introduction

Since the proposal of C-RAN [1-3] in 2009, China Mobile (CMCC) has been committed to developing various kinds of proof-of-concept (PoC), test-beds, and field trials to demonstrate C-RAN's benefits and verify the key enabling technologies. This chapter gives a comprehensive introduction to these activities. In particular, we will demonstrate not only the feasibility and reliability of wavelength-division-multiplexing (WDM)-based fronthaul (FH) solutions but also how a noticeable coordinated multiple-points (CoMP) gain can be achieved with the C-RAN architecture. In addition, a virtualized C-RAN system is elaborated, including the design principles, the architecture, and the field trial results.

Field-Trial Verification of FH Solutions

19.2.1 Centralization Field Trials in 2G and 3G Networks

The first step toward C-RAN was baseband unit (BBU) centralization which is relatively easy to implement and can be tested with the existing 2G, 3G, and 4G systems. In the past few years, extensive field trials have been carried out in more than 10 cities in China using commercial 2G, 3G, and pre-commercial TD-LTE networks with different centralization scales. The main objective of C-RAN deployment in 2G and 3G is to demonstrate the deployment benefits of centralization, including accelerated site construction and reduced power consumption. For example, one trial took place in the city of Changchun where 506 2G BSs in five counties were upgraded to a C-RAN-type architecture centralized in several sites. In the largest of these, 21 BSs were aggregated to support 101 RRUs with a total of 312 carriers. It was observed that power consumption was reduced by 41% owing to shared air-conditioning. In addition, system performance in terms of the call-drop rate as well as the downlink data rate was enhanced using multiple RRU-co-cell technologies. For the results and benefits from using centralization in 2G and 3G trials, the reader is referred to [4]. When it comes to TD-LTE, centralization becomes more challenging owing to the high data rate in the FH connection. For example, the data rate of the most widely used FH interface in the industry, the common public radio interface (CPRI), could be as high as 9.8 Gb/s for an TD-LTE carrier with a 20 MHz bandwidth and eight antennas.

Introduction

The trend of increased centralization holds the potential to transform mobile networks in two ways. First, centralization enables the exploitation of common channel knowledge, which in turn allows for significant improvements in the performance of a communication channel by, for instance, performing the joint transmission and reception of signals or allocating resources jointly amongst adjacent cells [1]. Second, centralized processing leverages the trend towards deploying mobile networks on low-cost commodity hardware that is running commodity or open-source software solutions. Deploying software-based implementations increases implementation flexibility, reduces service-creation time, and enables the flexible usage of processing resources through virtualization. In this chapter we use the term Cloud-RAN (C-RAN) to refer to a flexible use of commodity solutions that combines gains in both the telecommunication and information technology domains.

Before implementing the protocol stack of a RAN on a cloud-computing platform, we must also take the required effort into account, e.g., commodity hardware is considered to be less performant and energy efficient than dedicated hardware such as ASIC, DSP, or FPGA. Furthermore, resource virtualization implies an overbooking of resources while satisfying joint resource requirements of all processed base stations (BSs), which is in contrast with fulfilling individual processing constraints at each BS. Centralized signal processing may further impose stringent requirements on the fronthaul network between a radio access point (RAP) and the data center.

So far, research in the area of Cloud-RAN has focused on the telecommunication domain, e.g., the applicability of joint processing approaches, gains from centralization, and optimal degrees of centralization under different side constraints. In this chapter the focus is on the impact of limiting and virtualizing the data processing resources on the communication rate, i.e., the quantitative coupling of the required computational resources and communication rates [2]. After introducing basic notation and definitions, we consider metrics and an analytical framework that allows one to determine the data processing demand Interestingly, the data processing requirements depend not only on the number of information bits but also to a large extent on the quality of a user's communication channel. In this chapter we discuss and quantify multi-user gains, which lower the requirements on the data processing resources to be provided.

Introduction

In 2008, as the specification for long-term evolution (LTE) Release 8 was frozen in the Third Generation Partner Project (3GPP), operators began to shift the network deployment focus to 4G. In 2009, the world's first commercial LTE network was launched by TeliaSonera in Norway and Sweden. As of today, there are several hundred LTE networks in operation, providing unprecedented user experiences to customers. Consequently, we are witnessing the recent mobile traffic explosion in the telecom industry. It is expected that by 2020 consumer Internet traffic will increase by a factor of over one thousand [1].

As operators roll out and expand 4G networks, more and more challenges arise. First, network deployment is becoming more and more difficult simply due to an insufficient number of equipment rooms. Traditional base stations (BSs) comprise either a co-located baseband unit (BBU) with a radio unit or a distributed BBU with a remote radio unit (RRU) connected via fiber. For either case, a separate equipment room with supporting facilities such as air conditioning is required in order for BS deployment. However, since the operating frequency of LTE is usually higher than that of 2G and 3G, the coverage of an LTE cell is smaller than that of a 2G or 3G cell. As a result, more LTE cells are needed to cover the same area, meaning that more equipment rooms are required. Unfortunately, this is increasingly difficult since available real estate is becoming scarcer and more expensive. Traditional deployment puts a lot of pressure on capital expenditure (CAPEX).

Second, in a society where people are promoting energy conservation and environment protection, power consumption has become a sensitive word and a major concern for operators. It is estimated that the carbon footprint of the ICT industry accounts for 2% of the global total, which is the same as that of the aviation industry. For the telecom industry, further analysis has shown that a large percentage of power consumption in mobile networks comes from radio access networks (RANs) [1, 2]. Take China Mobile's networks, for example. The largest mobile network in the world consumed over 14 billion kWh of energy in 2012 in its network of 1.1 million base stations. It can be seen that saving energy in RANs could directly lower the operating expense (OPEX) of the network.

Introduction

The C-RAN architecture relies on fronthaul links to connect each remote radio head (RRH) to the managing baseband unit (BBU). In particular, for the uplink, the fronthaul links allow the RRHs to convey their respective received signals, either in analog format or in the form of digitized baseband samples, to the BBU. For the downlink, the BBU transfers the radio signal that each RRH is to transmit on the radio interface, in analog or digital format, on the fronthaul links to the RRHs. It is this transfer of radio or baseband signals that makes possible the virtualization of the baseband and higherlayer functions of the (RRHs) at the BBU, which defines the C-RAN architecture. The analog transport solution is typically implemented by means of radio-over-fiber (see, e.g., [1]) but solutions based on copper LAN cables are also available [2]. In contrast, the digital transmission of baseband, or IQ, samples is currently carried out by following the common public radio interface (CPRI) specification [3]. This ideally requires fiber optic fronthaul links, although practical constraints motivate the development of wireless-based digital fronthauling [4]. The digital approach seems to have attracted the most interest owing to the traditional advantages of digital solutions, their including resilience to noise and to hardware impairments as well as flexibility in the transport options (see, e.g., [5]). Furthermore, the connection between an RRH and the BBU may be direct, i.e., single-hop, or it may take place over a cascade of fronthaul links, i.e., be multi-hop, as illustrated in Fig. 8.1.

In this chapter we provide an overview of the state of the art on the problem of transporting digitized IQ baseband signals on the fronthaul links. As mentioned, the current de facto standard that defines analog-to-digital processing and transport options is provided by the common public radio interface (CPRI) specification [3]. This specification is widely understood to be unsuitable for the large-scale implementation of C-RAN owing to its significant fronthaul bit rate requirements under common operating conditions. As an example, as reported in [5], the bit rate needed for an LTE base station that serves three cell sectors with carrier aggregation over five carriers and two receive antennas exceeds even the 10 Gbits/s provided by standard fiber optic links.

Cloud radio access network (C-RAN) architecture offers two key advantages as compared with traditional radio access networks (RANs) from the physical-layer transmission point of view. First, the centralization and virtualization of RANs allow the coordination of base stations (BSs) across a large geographic area, thereby enabling coordinated physical-layer resource allocation across the BSs. The physical-layer resources here refer to the frequency, time, and spatial dimensions that can be utilized by radio transmission. Second, and more importantly, the C-RAN architecture also opens up the possibility of the joint transmission and joint reception of user signals across multiple BSs, thereby fundamentally addressing the issue of inter-cell interference. As interference is the main bottleneck in modern densely deployed wireless networks, the C-RAN architecture offers significant advantages in that it provides the possibility of interference mitigation leading to performance enhancement without the need for additional site and bandwidth acquisition.

This chapter provides an optimization framework for cooperative beamforming and resource allocation in C-RANs. We begin by identifying frequency, time, and spatial resources in wireless cellular networks and defining the overall spectrum allocation, scheduling, and beamforming problem in a cooperative network. The chapter then provides a network model for the C-RAN architecture and illustrates typical network objective functions and constraints for network utility maximization. A key characteristic of the C-RAN architecture is that the fronthaul connections between the cloud and the BSs may have limited capacities. One of the main goals of this chapter is to illustrate the impact of limited fronthaul capacity on the cooperative beamforming and resource allocation in C-RANs.

The chapter explores the optimization of the design variables associated with CRANs, depending on the transmission strategies at the cooperative BSs. For the uplink C-RAN, we illustrate compress-forward as the main strategy at the BSs and focus on the impact of the choice of quantization noise levels at the BSs and possible joint transmit optimization strategies. For the downlink C-RAN, we compare a compression-based strategy and a data-sharing strategy and illustrate the problem formulation and solution strategy in both cases. Throughout the chapter, key optimization techniques for solving resource-allocation problems in C-RANs are presented.