Table 1: Different types of Optical fronthaul deployment technology with their pros and cons 
- A. Optical Fronthaul
Early deployment of C-RAN in China, Japan, South Korea use optical fiber as the only fronthaul candidate. It is mainly due to its high BW and low latency. According to Mavrakis , optical fronthaul deployment can follow different technologies such as dedicated fiber technology, Passive CWDM technology, Active WDM technology, and Passive Optical Networking (PON) technology. All the above technology uses a mix of passive and active optical equipment. The advantages and disadvantages of each technique are listed in Table 1.
- B. Wireless Fronthaul
Wireless fronthaul seems to be an alternate of optical fronthaul. It is cheap. RRH can be installed anywhere due to no physical wire connection, lightweight, less space requirement, and less construction required. So, it is easy to install either in dense urban areas or any hilly region. It takes less time for setup. Due to its many advantages now, many vendors like EBlink, Cablefree are mainly focused on the development of wireless fronthaul. According to one of the vendor Cablefree , wireless latency is less compared to optical fiber due to the following reasons: 1) Optical fiber paths are not in a straight line, 2) Wireless propagation is faster (around 40%) than optical fiber due to lower refractive index of air than fiber.
- C. Ethernet Fronthaul
The use of Ethernet for fronthaul is a recent development in C-RAN. For the successful implementation of Ethernet fronthaul, many projects such as CERN’s white rabbit project , European Union’s Horizon 2020 project (iCIRRUS)  are going on. Research groups like IEEE have already formed its task force named as IEEE1904.3 , IEEE1588  and IEEE Standard Association (SA)1914  for encapsulation of CPRI over Ethernet, its synchronization and for standardization. Ethernet fronthaul has the following benefits. Ethernet makes virtualization of C-RAN cheaper and easier as SDN is already developed for IP networks. Due to the use of routers/switches, statistical multiplexing gain is possible. Monitoring, fault finding, managing is possible due to Operation, Administration and Maintenance (OAM) operation over Ethernet. It is cost-effective and can use the existing Ethernet network.
5 Technical Challenges for C-RAN Fronthaul
In this section, the technical challenges of existing fronthaul candidates in terms of throughput, latency, and jitter requirement, their deployment issues, availability, and the future scope are mainly highlighted.
- A. Optical Fronthaul
Optical fronthaul follows a physical wired connection. So for dense urban areas or hilly areas, it is not suitable for deployment of the optical fiber due to its inflexibility. It is also very unsuitable for disaster-prone regions. Implementation of optical fronthaul is a time-consuming process. One major disadvantage of optical fronthaul is its high cost. So for small distance areas, optical fronthaul is unsuitable. It is fully unsuitable for some regions like Africa, Southeast Asia, and some European countries which will not prefer to bear fiber deployment cost.
- B. Wireless Fronthaul
CPRI fronthaul capacity requirements are very high. It is difficult for a wireless link to achieve the growing demands of CPRI due to its constraints in terms of spectrum and distance if mm-wave is taken into account. Although vendors like EBlink have already proposed devices for wireless fronthaul, which can support up to 7.5Gbps CPRI data rate for 70MHz channel BW. While using the millimeter band, the wireless fronthaul can support up to 2.5Gbps for 500MHz channel BW . In 5G, due to the use of a massive MIMO antenna system, the CPRI data rate requirement will be more than 100Gbps. It seems to be difficult for wireless fronthaul to support such a huge data rate.
- C. Ethernet Fronthaul
For the existence of Ethernet fronthaul, it must meet the two major stringent performance requirements of CPRI, i.e., a) delay should be within $100 \mu s$ b) jitter should be within 65 ns. But standard Ethernet is unable to meet the performance limit. In standard Ethernet, although the delay is under control, jitter is very high (400 ns in the worst case) . To carry time-sensitive traffic in Ethernet, IEEE has proposed two new enhancement technique named as IEEE802.1 Qbu (used frame preemption) and IEEE802.1 Qbv (used scheduling). In IEEE802.1 Qbu frame with lower priority is preempted by frame with higher priority, and a frame can be preempted several times.
Similarly, in 802.1 Qbv, a frame gets scheduled to a transmission gate, which is either an open or closed state. Timing signals control transmission gates. The experimental result says Ethernet with frame preemption can never achieve jitter requirement (65 ns) either with background traffic (packet injected by nodes) or only with CPRI traffic . By using IEEE802.1 Qbv (Packet scheduling), it is possible altogether to remove jitter in case of without background traffic (only CPRI traffic). In the case of background traffic, jitter is below than CPRI standard. Jitter depends on the packet size and the slot size. By selecting a global scheduling instead of local scheduling, jitter can be removed entirely for the slot size of 1.5 times of Packet Transmission Time (PTT). Maximum delay achieved in the case of Ethernet with preemption for without background traffic is below to the CPRI standards, but it can’t reach jitter limit. In the case of Ethernet with scheduled traffic, the maximum delay is lower than $100 \mu s$. So it is concluded that only IEEE802.1 Qbv is sufficient to make Ethernet fronthaul feasible for CPRI standards. The physical deployment of Ethernet fronthaul is still far away. All the above results and conclusions are simulation-based. It can not guarantee that Ethernet fronthaul can work in 5G.
The above three CPRI-based fronthaul is working well for 4G and the traditional network like 2G, 3G, but it is unsuitable for the next-generation network, like 5G. It is due to the following reasons.
A CPRI-based fronthaul carries I/Q sample data between BBU and RRH. This sampling data have a direct relationship with the number of antennas used at the RRH side. So the required fronthaul capacity, which is also known as the CPRI data rate, increases as the number of antenna increases. For example, A 2$\times$2 and 4$\times$4 antenna configuration with two sectors for 20MHz BW, the required CPRI rate is 4.9Gbps and 9.83Gbps respectively . The fronthaul data rate is directly proportional to the number of sectoring and the number of antennas used in one sector. For each sector, the data rate gets double. For 5G, due to the use of 64 or even 128 antennas at RRH, the required data rate will become the bottleneck.
CPRI follows a constant data rate due to synchronous digital hierarchy-based (SDH-based) transmission mode . So there is a chance of low utilization efficiency during the night time or maybe at non-official hours. Even for a dense urban area tidal wave effect is not noticeable at every time.
Under CPRI based connection, there is a fixed one to one correspondence between RRH and BBU pool. So during an emergency, one RRH cannot be automatically switched to another BBU pool by automatic routing. This kind of flexibility is missing.
Statistical multiplexing is absent in CPRI based communication. It is due to the distribution of time-domain radio waveform samples over a fronthaul network.
Missing of Operation, Administration, and Maintenance (OAM) operation.