Today SON functionalities are considered an integral part of future RANs. They provide a means to simplify network management tasks, to reduce operational costs, and to enhance network performance and profitability. The inclusion of SON functionalities in the 3GPP Long Term Evolution (LTE) standard of beyond 3G networks is illustrative of the new trend in autonomic networking.
Auto-tuning is a self-optimizing function that aims at dynamically optimizing network and Radio Resource Management (RRM) parameters. Auto-tuning has been among the first SON functionalities studied in RANs. Parameter adaptation allows the network to adapt itself to traffic variations and modifications in its operation conditions, such as the introduction of a new service or a change in the weather and the propagation conditions. Hence a better use of the deployed infrastructure and of the scarce radio resources can be achieved when implementing auto-tuning. Self-optimizing networks have been extensively studied in the last decade for different RAN technologies. More recently, the scope of SON has been enlarged and covers more mechanisms of network operation
Planning: Network planning includes the assessment of the required infrastructure in greenfield or in densification scenarios in order to meet some coverage and capacity targets. Interfacing network databases of network configuration and of the monitored performance data to a planning tool can simplify the process of network planning. Identifying the needs, the required resources, and their location can be done automatically using automatic planning tools that are interfaced to the network.
Deployment: When deploying network elements such as a new base station, the material should be correctly configured. Self-configuration deals with the automatic installation and parameters’ configuration of new network elements.
Maintenance: Self network testing, monitoring, diagnosis, and healing are part of the autonomic maintenance process.
To implement self-operating functions such as self-configuration, self-optimization, or self-healing, self-awareness capability of network elements at different levels is required. Appropriate information should transit between network elements and should be processed. In self-optimizing networks, a network element may need to learn the best action it should perform in a given network state, which is defined by its environment. The entity capable of performing actions that modify the network (i.e., utilized resource, parameters, or RRM algorithm) and that has learning capabilities is called an agent. The network may have many agents operating simultaneously, at each base station, for example. The agents can exchange information and can act in a cooperative manner. The agents’ capability of sensing and reacting intelligently to the environment and of performing tasks can be referred to as cognitive mechanisms.
Certain self-optimizing network functionalities are associated with RRM functions, and their implementation depends on the utilized Radio Access Technology (RAT).
SON functionalities encompass different aspects of network operation. Their implementation can considerably alleviate operational complexity and speed up operational processes. SON aims at improving both network management and optimization.
The illustration below presents the RAN architecture with some of the SON functionalities. Self-optimization, healing, or diagnosis functionalities could be implemented within the Operation and Maintenance Centre (OMC) or could be located elsewhere, that is, in a dedicated SON entity.
For this reason the corresponding blocks are partially situated in the OMC. The Data and Configuration Management block includes two functions: reporting network configurations to the OMC that can be used, for example, by the planning tool; and data management that includes data collection and processing. Certain data management functions such as filtering are performed outside the OMC. Locating the planning tool inside the OMC aims at interfacing the tool with the network and at updating it with precise network configurations and other performance and QoS data. Furthermore, one can interface different SON functionalities directly with the planning tool. Two options can be envisaged for implementing self-optimizing functions and, in particular, the auto-tuning of RRM parameters: online and offline autotuning. Online auto-tuning operates as a control process with a time resolution varying from seconds to minutes in which RRM parameters are tuned in a control loop. The auto-tuning in this case can be seen as an advanced RRM functionality (see the section Control and Learning Techniques in SON). Offline auto-tuning operates in a time resolution varying from hours to days and can be seen as an optimization process.
Self-Configuration
The self-configuration process is defined in the LTE standard as “the process where newly deployed nodes are configured by automatic installation procedures to get the necessary basic configuration for system operation”. Following the deployment of a new base station (or eNB in the LTE technology), different parameters should be configured and functions should be executed such as:
- Hardware and software installation
- Transport network setup (IP addresses, setup QoS parameters and interfaces, etc.) Authentication
- Automatic Neighbor Discovery (AND)
- Radio parameter configuration (handover, selection–reselection, power settings, etc.)
- Remote testing
Initial parameter settings can later be improved in the self-optimization process. An example is the AND, or the Automatic Neighbor Relation Function (ANRF), in the LTE standard. The ANRF function is based on the capability of a mobile to send to its serving eNB the Physical Cell Identity (PCI) of the cells it senses. The serving cell can then request from the mobile to send the Global Cell Identity (GCI) of the sensed eNB, and once it receives this information,\ it can decide to add this cell to its neighboring list. In a self-optimization process, the neighboring cell list can be updated to follow the evolution of the network.
SELF OPTIMIZATION
The term self-optimizing network refers to the network’s capability to dynamically adapt itself to changes such as traffic variations or other changes in the operating condition. In the LTE standard, self-optimization process is defined as “the process where UE (user equipment) and eNB measurements and performance measurements are used to auto-tune the network. This process works in operational state.” Auto-tuning allows full exploitation of the deployed resources, an improved success rate of RRM procedures, and increased spectral efficiency and network performance. Furthermore, the burden on network management is alleviated. Use cases of self-optimization in LTE are described in the section Overview of SON in RANs. Studies on self-optimizing networks have shown that it is particularly useful for balancing traffic between base stations of a given RAT or between base stations in a heterogeneous network. Traffic balancing can be performed by auto-tuning handover or selection– reselection parameters, and can often bring about important capacity gains. A second approach for load balancing is the Dynamic Spectrum Allocation (DS-Allocation). The idea of DS-Allocation is to share frequency bandwidth between different RANs (capacity pooling). When a RAN needs extra spectrum resources, the DS-Allocation process can locally reallocate these resources from a less loaded RAN in the overlapping area. The DS-Allocation requires first an evaluation of the required spectrum resources according to the traffic demand; then a suitable spectrum subband allocation is performed that maximizes spectral efficiency. A third approach for sharing resources between different RANs is the Dynamic Spectrum Access (DS-Access). The enabling technology to carry out DS-Access is Cognitive Radio technology. The definition of cognitive radio is a “radio (i.e., radio system) that can change its transmitter parameters based on interaction with its environment.”
Self-Diagnosis
Network diagnosis refers to the process of identifying the cause of faults in the network. A cause could be a hardware failure such as a broken baseband card in a base station or a bad parameter value, that is, transmission power, antenna tilt, or a control parameter such as a RRM parameter. A fault in the network can generate multiple alarms that make it hard to identify its cause. In this case the alarm correlation process may be required to access the root source. A fault may trigger no alarms but result in low performance, poor QoS, and failure of RRM procedures. Often the diagnosis of fault cause requires processing information comprising both alarms and performance indicators. The OMC gathers alarms, counters, and more evolved indicators from the network. A self-diagnosis function can be implemented as an entity within the OMC that can access and process information and perform statistical inference.
Self-Healing
Once the cause of a problem has been identified, the self-healing can be applied. If the fault cause is identified with a high degree of confidence (or probability), a correction action can be triggered and be reported to the management entity. If the degree of confidence is not too high, it is proposed to the management entity and is executed only after receiving a manual validation.
Self-Protecting
Security in RANs concerns the ensemble of protection operations such as access control, authentication, and encryption services. The security system monitors the security-relevant contextual input and can reconfigure itself to efficiently respond to dynamic security context. The self-protecting system can modify the cryptographic level, change the strength of the authentication scheme, install system patches to defeat new attacks, and so on.
