In the recent literature, it can be clearly noticed that there is a growing interest in virtu- alizing networks’ wireless resources [80]. For example, a wireless resource virtualization scheme for LTE evolved Node-B (eNB) is investigated in [62]. The scheme allows MNOs
Physical Network BPU Pool Fronthaul network Hypervisor MNO-1 Virtual Network MNO-2 Virtual Network RRH Virtual RRH for MNO1 Virtual RRH for MNO2
Figure 5.1: Virtualized C-RAN shared between two MNOs.
to implement different scheduling policies. However, the scheme did not consider network- wide virtualization nor coordination between interfering cell zones to prevent ICI. In addi- tion, the scheme suffers from high complexity because it requires solving two optimization problems to maintain isolation between MNOs. In the first optimization process, the re- sources of each MNO are allocated to their users. The allocation results of problem one are then fed to the second problem as constraints such that the throughput each user obtains is equal or greater than the throughput achieved when sharing is not considered.
An LTE air interface virtualization scheme is proposed in [67], where a hypervisor is added on top of the physical resources. The hypervisor is responsible for virtualizing the eNB into a number of virtual eNBs that can be used by different MNOs. It is shown that more capacity can be achieved by sharing spectrum resources between different MNOs. However, the scheme does not provide optimal solutions nor manage ICI. Furthermore, the instantaneous channel quality of users is not considered in the scheduling decisions, which limits multiuser diversity gain.
the hypervisor manages the sharing process of multiple eNBs among multiple MNOs. Nev- ertheless, only fixed resource allocation across BSs is considered. The load is balanced be- tween multiple BSs by moving users from high-traffic cells to low-traffic cells. However, transferring users across cells increases handover overhead, and may degrade the system capacity since users may be transferred to BSs further away, which would reduce the qual- ity of the wireless link.
Another framework for wireless network virtualization that separates service providers from a network operator is reported in [83]. The service providers (SPs) are responsible for QoS management, while the network operator is responsible for spectrum management. The interaction between SPs and the network operator is modeled as a stochastic game reg- ulated by the network operator. The role of the SPs is to compete for wireless resources for each subscribed user.
Utility-based resource provisioning scheme for WRV with massive MIMO is inves- tigated in [84]. A single BS equipped with a large number of antennas serves users of different service providers. The problem is formulated as a combinatorial optimization problem of high computational complexity. Consequently, a low-complexity solution for the combinatorial problem is derived by linear programming relaxation.
In multi-cell systems, the same frequency bands can be assigned to users in different cells, which is referred to as the frequency-reuse (FR) principle, which is used to increase both coverage and capacity. However, to minimize ICI, cells that use the same frequency bands should be separated by a sufficient distance. Several ICIC techniques have been proposed for multi-cell systems as described in [85] and the references listed therein. The most promising is the fractional FR (FFR), which is adopted by 3GPP LTE [86].
The performance of FFR has been extensively studied for traditional cellular net- works [52, 87, 88]. For C-RAN architecture, a dynamic FR scheme based on FFR is pro- posed in [89]. The wireless resources are assigned to cell zones using a graph-coloring approach. Each color represents a certain segment of bandwidth. To minimize ICI, differ- ent colors should be assigned to interfering zones.
A dynamic interference coordination scheme for downlink multi-cell systems is pre- sented in [71]. The allocation problem is divided into two sub-problems, one at the BS level and the other at the central controller. It is assumed that BSs are able to communicate
with each other using an X2 interface. At the BS level, each sector potentially allocates bandwidth chunks to its connected users. Then, each sector sends a request to the central controller. The request specifies a list of bandwidth chunks to be restricted at the dominant interfering zones. Conflicting requests are resolved by the central controller, which sends a refined list of chunks that should be restricted to each sector.
Minimizing network power consumption of C-RAN is investigated in [90], where the power consumption of the transport network and RRHs is considered. The authors as- sume that transport links and RRHs can support sleep mode. The problem is formulated as a joint RRH selection and power minimization beamforming problem. The network power consumption is reduced by minimizing the number of active RRHs and reducing their transmit power subject to QoS constraints. Through simulations, the authors show that the network power consumption can be notably reduced. The performance of CoMP transmission schemes in a C-RAN architecture for LTE-A Heterogeneous networks is stud- ied in [91]. With C-RAN architecture, a larger number of RRH can be considered in CoMP transmission, which improves the transmission performance.