The future of Radio Mobile Network testing and optimization. 4G In Vitro: Closing the gap between simulation and live experiment

"Experiments vary greatly in their goal and scale, but always rely on repeatable procedure and logical analysis of the results."  Wikipedia

 

In legacy systems, the use of a static radio planning followed by field trials was enough to meet the the Mobile Network Operators (MNO) performance requirements. 4G LTE and LTE-A materializes a move towards very dynamic and flexible networks. The resulting complexity requires new tools and methods to enable MNOs to fully exploit the performance potential of LTE technology and to reduce the deployment cost.
This white paper describes a new generation of testing and optimization tools, aiming at reducing the gap between simulation and reality, in order to accelerate deployments while reducing heavy OPEX and CAPEX costs, by minimizing the number of drive tests required on the field.
4G IN VITRO implements the full network in the lab. The main challenge is to demonstrate that the field environment can be reproduced accurately within the lab, while new tools and methods enable the operator to extract the KPIs as expected for performance, coverage, and capacity.
Today mobile network deployment and optimization are essentially driven by RF planning simulation phase followed by drive testing in a live experiment phase as shown on Figure 1.
RF planning uses advanced RF conditions and traffic models but it treats network elements such as mobile phones and Base station as  abstract models. Neither the protocol stacks nor the scheduling algorithm are included in these simulations, and neither the network elements nor the dynamicity is tested in that environment.
On the opposite side, live experiment (field trial) require high deployments cost and yield very low reproducibility capabilities. Usually 3 to 5 drive tests per site are conducted for different validation and optimisation needs.
The process of selecting the right network elements such as eNodeBs is usually done on specifications or based on live demonstration. There are no existing tools that can provide benchmarking capabilities with truely reproducible RF and load conditions giving a fair comparison between different vendors.


Figure 1 : Network deployment phases

 

Based on some research works [1]; we define three dimensions to evaluate the performance of a radio network testing tool.
The first dimension is the realism: radio environment, topology, and network elements behavior (UEs, eNodeB).
The second dimension is the deployment cost of the experiment: hardware provision, installation, configuration, software installation, and application launch.
The third one is the capability to reproduce the test and to control it: replay a scenario, modify the configuration, and monitor the running status.
As illustrated in Table 1, the realism of live experiment is very good compared to simulation while its deployment cost, control and reproducibility are bad.


Table 1:  Live experiment versus Simulation in term of realism, deployment cost and control [1]

The challenge here is to combine in the same testing procedure high realism, high reproducibility and low deployment cost.
Today, new cloud based architectures provide sufficient computational resource, testing control and flexibility to manage this paradigm.
Furthermore, simulation tools such as RF planning tools use software models, hardware models, and radio environment models.
Here we propose 4G IN VITRO, an emulation tool that integrates the real eNodeB equipment (real hardware) and runs the real protocol stack on UEs and eNodeBs (real software). The full network is emulated within the lab in order to manage reproducibility and testing control.
A significant advantage of running emulation connected to real eNodeBs (real hardware and software) is the ability to play the test with real running schedulers which are the core function of the eNodeB.


Table 2: Real Hardware versus Hardware model

 

4G In Vitro aims at reproducing the full field network behavior in the lab (Figure 2).
A key innovative concept here is the coupling of the traditional load emulation tool with radio planning tool in a unique system. This way, 4G IN VITRO can be configured with a specific field area conditions by importing the detailed coverage predictions and site locations of the chosen area from the RF planning outputs. Figure 3illustrates the way 4G IN VITRO is configured with the Toronto city area.


Figure 2 : Emulation tool with real hardware, real protocol stacks, and dynamic radio environment model.

 

The radio environment is fully emulated in 4G IN VITRO: including fast fading, path loss, mobility and interference.
The signal level (RSRP) and signal quality (RSRQ) grids of the target zone are imported from the radio planning tool (Figure 3), meaning that realistic pathloss and handover measurements are available. Appropriate traffic models are also used to position the UEs. Each group of UEs is also associated with a fast fading model based on its speed and location.
Interference is the major limiting factor in a mono frequency network such as LTE. The 3GPP standard suggests dynamic management of interference.  4G IN VITRO is the only tool with capability to stimulate accurately the eNodeB ICIC (InterCell Interference Coordination) algorithms.  For each UE, the UL and DL levels of interference are emulated dynamically using the actual scheduling and load of adjacent cells and actual location of the UEs.
On top of the radio environment emulation and traffic generation, 4G IN VITRO provides the capability to collect multiple layers of counters, aggregates them and constructs the expected KPIs to evaluate network performance.
Finally, 4G IN VITRO emulates the UE behavior, meaning that it modelizes specific vendor’s characteristics on modem chipset, OS or applications.


Figure 3 : Toronto city RSRP and RSRQ levels import from radio planning tool into 4G IN VITRO