Home > 5G Network > Fronthaul Design In Cloud Radio Access Networks: A Survey

Fronthaul Design In Cloud Radio Access Networks: A Survey

In 2016 a colossal work was done on the architecture of C-RAN. In [24], a novel model has been created to deal with high capacity data transfer in urban areas. In urban areas, there is a huge gap in data transfer rates between the business area and the residential area. In this work, a set of positions of BS is taken as input and come up with optimal BBU positions as output. To build the model, delay, load balance, and capacity requirement are considered as parameters. This analytical model is composed of four components, namely, RRH proliferation, BBU positioning, load balancing, and BBU dimensioning. The simulation results are promising and successfully evaluate the best deployment configuration from the operators’ perspective. The result also shows that in that way, the resources in the BBU pool can be decreased by up to 25% in their peak loads.

At the same time, in [25], critical technologies used in C-RAN are discussed. In this work, it was shown that traditional hexagonal topology does not fit with the current C-RAN deployment model. As the introduction of the microcell, picocell, femtocell in a multitier and ultra-dense network, HetNet topology cannot be depicted as a hexagonal grid because BS or RRHs are random in this case. Moreover, the ON/OFF facility in femtocell makes it more randomized. So, a frameless model is necessary. This model is called as Frameless Network Architecture (FNA). The stochastic geometry method, with a Poisson Point Process (PPP) model, has been proposed for this type of dynamic network. One very recent model, Ginibre Point Process (GPP), has been submitted for depicting this deployment topology. In FNA, there is no need for handover. The serving set will be updated automatically to ensure the QoS requirement.

In [26], a novel idea for the betterment of the performance of 5G HetNets and facilitate offloading management is presented. Their main aim is to create cloud RRH. That is another level of abstraction over the physical RRH. The physical RRHs are like femtocell, and many femtocells are connected to a small cell, that is the cloud RRH. The cloud RRH has CPU, memory, and storage space for intermediate processing. The physical RRH and the cloud RRH is connected by radio wireless link. The cloud RRHs are interconnected to each other through Z-interface. For resource and interference management, one RRH from each small cell will be connected to the BBU pool through CPRI Interface. In this way, CAPEX will be reduced. The result shows that Quality of Service (QoS) is enhanced compared to conventional C-RAN even though the authors have come up with some new approach of C-RAN architecture. Any efficient offloading algorithm is not mentioned. At the same time, how mobility and inter-cell interference can be handled in this type of scenario was also not discussed in this work.

In [27], an improved version of optical-access-enabled C-RAN architecture is proposed. To enhance signal qualities, CoMP and Train Relay Station plays a vital role in railway communication. For that, BS coordination is very much required. Now during the BS coordination, an excessive data exchange has to be done, thus occupying excessive bandwidth, and that will affect the backhaul transmission negatively. To get rid of that situation, Cross-BS fronthaul link (CFL) comes into the picture. In this case, even if two RRHs belong to different BSs trying to send data to a single BBU, one of them can use CFL to transmit the data directly to any one of the two BBUs. As a result of both traffic, i.e., the fronthaul and backhaul traffic can be sent simultaneously; hence, the High-Speed Railway (HSR) communication is improved.

Though tremendous advantages of C-RAN can be seen nowadays from data rate to the total cost of ownership, multicell coordination is one of the biggest hurdles. At the same time, as the fronthaul is not ideal, latency and performance of the system can be affected by the functional split as well. In [28], the authors have investigated the effect of coordination algorithms like coordinated scheduling, dynamic point selection, and uplink joint reception for a given fronthaul latency. It was shown that the coordination performance in fully centralized C-RAN is no better than fully distributed or partially distributed C-RAN. That means the architectural evolution towards decentralization is a correct way to combat the vast data processing jargon. Furthermore, pooling gain is also being examined in this work.

Due to Network Function Virtualization (NFV) and Softwarization in C-RAN, the performance is promising. But as this whole system is becoming very complicated, switching and aggregation concept in fronthaul must come into the picture. At the same time, to combat the limited fronthaul capacity constraints, the functional split is becoming indispensable. In [29], the problem of a functional split is analyzed, and a solution approached is also given. The main problems of functional split or flexible centralization are 1. RRHs become more complex, that means expensive 2. It reduces the opportunities for multiplexing gains, coordinated signal processing, and interference management. But at the same time, it reduces the fronthaul consumption in the order of magnitude. So, there is a trade-off between the complexity of RRH and fronthaul consumption. Now to transmit data through CPRI or OBSAI, packetization has to be done. Received samples from ME is first packetized in RRH, then that packets are sent to the BBU pool, de-packetization is done there to get the samples back. Finally, all baseband processing is done. Now the question is that at what point packetization should be done. The analysis shows that if packetization is done immediately after getting the samples, then FH consumption is very high. If packetization is done after CP removal and FFT conversion, then the consumption will be low. But in that case (de)-packetization latency will be higher. This work examined the most suitable packetization method, taking packetization latency and FH overload into consideration so that HARQ deadlines are met. But this is very situation-specific. The multiplexing gain for different UE densities and functional splits are analyzed, and a conclusion is drawn on the maximum number of supported RRHs for the best packetization method.

Pages ( 4 of 12 ): « Previous123 4 56 ... 12Next »

Leave a Comment:

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.