Now, as the baseband modules are mostly on the cloud and that too can be managed, a new business model comes into the picture. Along with infrastructure, platform, and software as a service model, a new concept called RAN as a service is a budding business model. In , using cloud computing and C-RAN, RAN as a service model is proposed. In this model, the challenges and their possible overcome strategies are discussed. RANaaS is a flexible model based on centralized processing and also capable of handling interference in a very dense network. In , architecture is also proposed to offer RAN as a service. The novel work in this paper proposed a simulation-based model to analyze the performance of the RANaaS model. But whether the architecture is feasible in terms of efficient resource allocation, computational complexity, and load balancing in a real scenario and on a large scale is yet to be decided as no emulation has been done yet.
Apart from the above-discussed problems, one big problem is the functional split. Functional split decides which processing block (e.g., source encoding, modulation/demodulation, multiplexing/demultiplexing, etc.) is to be placed at the RRH side and which is to be placed at BBU pool. If most of the processing blocks are placed in RRH, then data transmission to the BBU will be less. Thus, the bandwidth requirement will be reduced, but then the RRH design will be very complicated, leads to an increase in the total cost of ownership (TCO). If all the processing blocks are at the BBU pool, then raw samples have to be transmitted to the BBU; thus, the fronthaul bandwidth requirement will be much higher. So, there is always a trade-off between RRH complexity and fronthaul capacity in C-RAN. In , constraints and outline applications of flexible RAN centralization is analyzed. In this work, the performance by PHY layer functional split and MAC layer functional split is examined. In the case of the PHY layer functional split spatial diversity can be fully exploited. By implementing advanced processing, inter-cell interference can also be mitigated, or sometimes it increases the overall capacity. But in this case, high capacity, and low latency backhaul is required. In the case of the MAC layer split, it enables coordinated RRM and scheduling. But then RRH will be complex and costly, as well. At the same time, joint decoding, CoMP, etc. will not work in this case. But the technical challenge of doing this type of splitting is not discussed in their work.
Further, in , a practical implementation of C-RAN architecture in the context of 5G systems, with functional split taking into consideration, is discussed. In this work, the benefits and challenges of C-RAN for 5G architecture is analyzed. How different hardware options impact the implementation of C-RAN is thoroughly investigated. Backward compatibility with 3GPP LTE is another concern of C-RAN. In , the compatibility with 3GPP in terms of latency and throughput is shown. The flexibility of the proposed architecture is described from a practical point of view.
As the fronthaul capacity constraint is one of the bottlenecks, a graph-based approach is proposed in  to deal with limited fronthaul capacity. In this model, nodes of the graph are the baseband functions, links are the information flows, and the weight of the links is computational and front hauling cost. Finally, the optimum location is found out to place the baseband functions by finding the optimum clustering scheme for graph nodes. The problem is solved using a genetic algorithm with a customized fitness function and mutation module.
Cloud computing is the core concept of C-RAN. Due to its centrality, fronthaul complexity increases with the number of active users and data transmission. So, to overcome these problems, a new edge computing scheme like fog computing is proposed. In , harmonization of C-RAN and fog computing is established. Though both have some advantages and limitations, their harmonization can make the whole system very efficient. Here the concept of local link, i.e., device-to-device link in close proximity and the remote link, is used at the time of downloading any content. The load in cloud and device for the highest performance is analyzed in terms of latency, throughput, and error.
A unified model is proposed in . Currently, various C-RAN models are proposed to explore the potential advantages of C-RANs. So a unified model of C-RAN for 5G is not straight forward.
Apart from RANaaS, C-RAN technology opens up many application fields where centralized processing of signal communication has to be done. As the resources are centralized and information from many different RRHs are coming to the baseband pool for processing, much joint processing for overall performance gain can be done efficiently. One of the recent works on this technology is the Indoor Distributed Antenna System (IDAS). In , an SDR based solution approach to tackle latency, fronthaul capacity resource control is discussed. At the same time, a practical implementation of the Indoor Distributed Antenna System (IDAS) is also presented by using the C-RAN platform. Their experimental study also discloses some answers like controlling the communication between RRH and BBU, managing BS resources in the server platform, impact of interface algorithm on the system. Traditional cloud computing resource sharing platform will not work in this case because of stringent timing constraints. How to use some advanced technology to deal with these timing constraints is still a research topic.