I expect many of you can remember those nerve jangling moments as the end of the six hour window to upload the new frequency plan approached and the network was about to go live again. Tools were developed to improve the success and reduce the duration of these outages but, as additional bands, node types and, with 3G, access technologies were deployed, it was clear that a new approach to managing such complex networks was required.
More automation is needed to determine and update the parameters used by each radio cell, including the frequency used, RF power level, neighbour lists, handover criteria and even whether to power down at quiet times of day. Manual processes in use today simply can’t scale up quickly enough to handle the much larger number of cells and interactions efficiently.
Introducing Self-Optimising Networks (SON)
The SON framework was initially developed as an essential component of LTE, but has also been extended to handle 2G and 3G.
Its objective is to improve user experience and control the OPEX associated with managing increasingly large and multifaceted networks. Through the SON framework the network is able to detect changes to the network itself, its environment and its utilisation and to determine and implement an appropriate course of action to optimise network performance.
This functionality will be spread across different network components
• inside the small/macrocell
• within the RAN vendors EMS (Element Management System) or NMS (Network Management System)
• externally in a separate OSS system.
In practice, it’s likely that a combination of these components with different capabilities requiring access to different datasets, and operating in different time periods. For example, the small cell reacting within seconds to a French Window being opened while the external planning system might remotely reconfigure the network between day and evening or power down some equipment in the night.
These complex networks with a mix of large and small cells are now commonly referred to as Heterogeneous Networks or HetNets. The original definition of the term HetNet which referred to a network containing radio nodes of different power classes (small cells and macrocells) has been extended to include a mix of access technologies ( eg 2G, 3G, LTE & Wi-Fi) and a range of frequency bands.
SON feature evolution
The initial SON use cases defined by NGMN focused on planning, deployment, optimisation and maintenance of the radio network. As LTE and Heterogeneous Networks mature, the scope of SON continues to evolve and the framework extends into areas such as energy management and backhaul configuration.
SON is standardised by 3GPP and is divided into 3 categories; self-configuration, self-optimisation and self-healing. Features are included from Release 8 onwards, with the initial focus on deployment related features such as eNB self-configuration, Automatic Neighbour Relation (ANR) and PCI Selection. Release 9 and 10 introduce functions to support expanding networks and include features such as Mobility Load Balancing (MLB) and Mobility Robustness Optimisation (MRO). Release 11 extends features such as ANR and MRO to support Inter-RAT uses cases while Release 12 introduces multi-vendor eNB self-configuration.
An overview of each 3GPP release is available here: 3GPP Release Descriptions
Vendors need to co-operate and co-ordinate their SON implementation
3GPP Release 8 also introduces the reference architecture for SON which identifies where SON algorithms may be executed. The reference architecture includes:
• NM-centralised and EM-centralised; which refer to the Network Management and Element Management nodes, respectively.
• Distributed; where decisions are made at the Network Element level
• Hybrid; where 2 or more of the above levels are included in the algorithm execution.
SON solutions are being developed by both the wireless infrastructure providers and by specialist Network Management Solution vendors. Most, if not all, specialist SON solution providers employ a centralised architecture using standardised information and proprietary algorithms. While this may increase latency or signalling load in certain procedures in comparison with a distributed architecture, one of the key benefits of this approach, particularly with respect to HetNets, is its multi-vendor capability.
The integrated SON solutions provided by the wireless infrastructure vendors can implement individual SON functions in the layer of the reference architecture which is deemed most beneficial. This, of course, is only applicable when managing the vendors’ own equipment and a centralised model needs to be applied when managing a multi-vendor network. There is broad debate across the industry as to whether certain SON functions, such as MRO, are best implemented in a centralised or a distributed manner. At this time, there is insufficient data to reach a conclusion as to whether one mode will outperform the other in live networks.
Promising results from early SON deployments
SON has been deployed as an integral component in many of the live LTE networks. Most of these networks are, however, still immature and have had limited opportunity to test the capabilities of SON beyond the self-configuration category. We have also seen rapid advancement in 3G SON solutions with a recent survey report by Infonetics (SON-and-Optimization-Strategies-Survey-Highlights) indicating that some operators have deployed SON as the key optimization tool in their 3G networks.
There are some reports on the benefit of SON in the self-optimisation category, such as this article posted in Q4 last year where AT&T highlighted SON as a key factor in reducing its dropped call rate by 10% and improving throughput speeds by a similar amount.
SON and Small Cells
Ease of integration with existing macro networks was a key requirement for femtocells from the outset and they therefore include several of the deployment and optimisation features associated with SON. Automatic configuration, neighbour generation, frequency selection and power adjustment are all included in femtocell solutions and their performance in indoor environments is well understood. What is less clear is how well these algorithms will perform in outdoor femtocell deployments where isolation from the macro network is reduced and closed subscriber groups are, for the most part, not used. This will be particularly challenging in 3G deployments where limited spectrum availability has often required femtocells to re-use, co-channel, spectrum from the macro network.
The “zero-touch” approach to configuration and management of femtocells is an ideal which we should aim to follow closely in outdoor small cell deployments. Although metrocells and picocells are being defined as network managed components, and are likely to have feature sets which are more closely aligned with the macro network than femtocell networks have been, the scale of deployment makes it essential that the SON criteria of self-configuration, self-optimisation and self-healing should be embedded within the small cell architecture.
Listening mode (where the 3G femtocell scans its surrounding environment to determine whether configuration changes are required) and a distributed architecture are key enablers to the degree to which SON uses cases have effectively been supported in a femto environment. As discussed previously, however, this distributed architecture is not well suited to a multi-vendor network. To overcome this challenge it is likely that a hybrid of distributed and centralised architectures will be required for outdoor small cells deployments.
SON is essential if we are to operate HetNets with a high level of user experience at an acceptable level of operational cost. Although initially developed as a key component of LTE, SON is being deployed in both 3G and LTE networks. While SON deployments are still in their infancy, positive improvements in the areas of self-configuration and self-optimisation have been reported.
For small cells it is important that as much SON functionality as possible be distributed to the Network Elements themselves to minimise latency and reduce load on the network management elements. This, however, needs to be balanced against a solution which enables multi-vendor operation.
For more detailed technical insight into SON, read a white paper on the topic from RANPLAN freely available from our download library