The density and complexity of mobile networks is increasing rapidly to cope with exponential evolution of user traffic, mostly driven by smartphones, connected applications and video streaming. LTE is an answer to these challenges with a new, very flexible transmission technology associated with many innovations. It enables heterogeneous network architectures mixing macro cells and small cells for extended coverage and capacity. This extended flexibility compared to previous network generations comes with increased complexity for the network operation activities, in a landscape under stronger and stronger cost reduction pressure.
SON is a solution to manage the complexity and decrease OPEX.
We can summarise it as a network operation automation technology which focuses on three main areas:
1. Self-Configuration functions: this is the ability for the network to re-configure itself automatically when nodes are added, deleted or modified. One example is ANR (Automatic Neighbour Relation), a feature that simplifies the reconfiguration process required when a new cell is added to a network.
2. Self-Optimisation functions: a recurring and automated process for the dynamic tuning of network parameters for optimal performances in changing conditions. For example, handling of traffic density migration linked to periodicity of business activities.
3. Self-Healing functions: automatic compensation of network nodes failures, to restore the service where it has been degraded. For example, self-healing can handle the network coverage loss in case of base station outage, by dynamic reconfiguration of adjacent healthy cells.
SON has been designed to decrease the manpower required for daily network operations, linked to new site deployments, handling of outages and fine-tuning of network parameters. Operational teams are under high pressure to deploy SON technologies, but they also need to keep the reliability of their networks, at least as the same level as before the SON era.
This is highly disruptive, as these operations have been managed manually for decades with reliable workflows, whereas SON maturity has not been proven yet. There is a legitimate strong resistance for swapping proven manual workflows by potentially divergent automated processes in a field where any error can lead to big revenue losses. In France, a recent 24-hour network outage has been estimated between EUR10 million and EUR20 million in repairs and compensation to customers.
Also, the growing scarcity of experienced resources in operational teams linked to the automation of daily tasks is a major concern: how to resume control of the system in manual mode in case of automation failure, when nobody owns the required know-how anymore?
We can compare the context of SON introduction to the one faced by the aeronautical industry about computerised flight control, mostly driven by the need to save weight and costs. The electric flight control technology has been used since 1958, but we had to wait for more than two decades to see in 1984 the generalisation of computerised flight command in the commercial planes industry. Extensive usage of simulation associated with more and more accurate test beds has been used for technology maturation and pilots training to achieve sufficient confidence before massive deployment.
In the following we will see how, as for the aeronautical industry, the mobile network ecosystem can take advantage of network emulation technologies to de-risk the deployment of SON.
De-risking SON deployment with network emulation
SON is a complex technology associating network measurements with sophisticated analysis and decision algorithms reproducing human reasoning.
The outputs of these algorithms are directly connected to operational levers of the network for a fully automated workflow. As networks are heterogeneous (multiple technologies, multiple vendors), SON systems must cope with the multiplicity of protocols, data formats and interfaces. They are looking more as labyrinthine system than as a Zen garden. Also, SON technology is so strategic that vendors put it under strong secret and sell black-boxes with very limited information on internal mechanisms.
This is where network emulation enters the game as a pragmatic solution to help all SON stakeholders to increase their trust and control over this new, promising technology.
We have seen previously that SON is tightly linked to dynamic aspects of the network with complex use cases involving multiple cells, multiple network nodes and SON equipments in front of many users with varying behaviours and sophisticated radio conditions. So, the validation and tuning of SON should be done at full system level with conditions as close as possible to the reality, including the trickiest part which is the radio environment.
Historically, tools available for mobile network testing were focused on only a few aspects of the complexity of the problem because of the limitation of available emulation technologies.
Typically, up to 2009, mobile network system testing tools where spread across the following main categories:
System level simulators used in R&D (network vendors or research labs) with very realistic models but absolutely no means to connect to real devices and nodes;
Radio channel emulators associated to real handsets to check system performances with a limited number of users (typically a few dozens of users);
Load and stress test tools able to generate heavy traffic (several hundreds of handsets) but with non realistic traffic, and no radio impairments emulations.
All these legacy techniques are still in use but cannot cover correctly the new test cases needed for modern telecom systems. The case of SON is critical. Proving its proper operation and stability requires using a complex system test bed composed of many different boxes coming from different vendors, with all features activated simultaneously, as in a live network. This is something that cannot be fully validated with legacy techniques.
Since 2009, we have seen test tool vendors coming up with breakthrough technologies able to reproduce very realistic conditions in a lab, including the radio path, under heavy traffic (several thousands of handsets). That has been possible thanks to the phenomenal increase of the CPU power and the ability to run in real-time radio propagation models on a per handset basis with a huge number of devices on chip computing platforms.
It is now possible to set up a test bed able to run the most critical SON use cases with real network nodes and heavy traffic.
As many SON mechanisms are based on radio measurements and cells radio coverage reconfiguration, the test bed should integrate a multi-channel radio path and interferences emulation component. It should be able to handle thousands of simultaneous different radio conditions (one for each handset/eNodeB couple). This technology can reproduce easily in the lab a very complex radio scenario so that the efficiency of SON mechanisms is evaluated with the accuracy required to trust SON systems.
SON technology is so strategic that vendors put it under strong secret and sell black-boxes with very limited information on internal mechanisms. This is where network emulation enters the game.
The image above shows a typical setup used to highlight a mobile network in the lab to assess the behaviour of the SON closed loop in front of different scenarios.
This test bed is composed of the following elements:
A network installed in a lab composed of all relevant real nodes and SON equipments. This example shows 3 eNodeBs with 3 sectors each, but other radio topologies can be considered easily
A handsets and radio channel emulator able to handle thousands of devices and reproducing very accurate traffic conditions down to interferences and fading effects
The emulator is configured for different scenarios needed to assess the performance and stability of the network and SON equipments such as:
Emulation of traffic variation throughout the day (high traffic at day time, low traffic by night) to check how SON reacts to achieve minimization of energy consumption
Emulation of interferences to assess the SON mechanisms in charge of coverage and capacity optimization
Emulation of cell outage to assess self-healing behaviour to detect and compensate cell failure
When a test campaign is running, QoE and QoS are automatically measured to build a picture of the level of end-to-end service experienced by users. Typically, a test which emulates an outage should show a temporary loss of service for some users followed by failover as soon as the SON system has detected and fixed the issue. When the network is reconfigured to handle the outage, the available capacity is lower because of the loss of one or several sectors; even if there is no coverage hole.
So, the users should experience a degradation of QoE if the capacity demand cannot be fully satisfied in this situation. This effect can be measured objectively through QoE estimation embedded in the test bed.
The following diagram shows a typical synthetic report generated by such a test bed when evaluating the self-healing feature.
In this diagram, we can understand how the self-healing feature of the network is configured:
1. At time 15, one or more sectors break down
2. We see many handsets losing connection to the network on the “No service” curve
3. The “VoLTE MOS”, “Video MOS” and “Web bitrate” are degrading because of radio coverage hole appearing around the faulty eNodeB
4. At time 45, the SON has detected the failure and reconfigured the healthy sectors to compensate for the eNodeB lost
5. Then, at time 60-90, we see “VoLTE MOS” going back near to its nominal value, meaning that voice communications are no longer significantly degraded
6. “Video MOS” is also recovering, but at a lower level because the network has lost capacity and it is configured to give priority to VoLTE
7. Web service is handled in best effort mode. As the demand is higher than the remaining capacity after the failure, it cannot recover to the initial level
As seen, the failover mode of the network causes performance degradation linked to weaker radio coverage and increasing interference for some users. These effects are fully emulated by the test bed for a valuable performance evaluation. By analysing the curves, the MNO can verify that the SON strategy is in line with its requirements. MNOs can tune network parameters to change the strategy, depending on commercial needs. For example, another MNO may choose a different strategy and put Video at higher priority than VoLTE. After having achieved the required behavior in the lab, the MNO can safely push the new network configuration to the live network.
SON is a promising technology that should help MNOs to save OPEX by decreasing the volume of manual tasks required for day to day network operation, and shall drive improvement in capacity, quality and network performance. However, SON deployment is slowed down by concerns about its effective maturity and by the organizational impacts on operational teams.
The diversity and complexity of SON systems implies sophisticated assessment logistic in order to achieve full confidence before effective deployment. Since major SON features are tightly linked to the radio path behavior, the assessment is difficult to achieve accurately and efficiently with standard techniques and tools.
Up to date emulation technology, involving realistic radio channel emulation with thousands of virtual handsets, is paving the way to help all stakeholders to successfully assess the SON benefits and deploy it in live networks:
1. SON providers can use network emulation to tune SON software and demonstrate the stability and benefits of their products to MNOs, with realistic and convincing traffic conditions
2. MNOs can validate proper operation of SON in a multi-vendor context with their own specific traffic model and network topology
3. Operational teams can train their employees on very realistic use cases, including simulation of emergency situations
With network emulation, the ecosystem can push the usage of SON securely; as flight simulation has helped the aeronautical industry to safely deploy next generation computerised auto-pilot technologies.
Frederic Rible is CTO for Mobipass at ERCOM.