Content uploaded by Jan M. Kelner
Author content
All content in this area was uploaded by Jan M. Kelner on Jul 03, 2023
Content may be subject to copyright.
Relationships Between QoS/QoE Metrics in Selected
Mobile Network Measurement Campaign
Dariusz Zmysłowski and Jan M. Kelner
Institute of Communications Systems, Faculty of Electronics, Military University
of Technology, Warsaw, Poland
dariusz.zmyslowski@wat.edu.pl
jan.kelner@wat.edu.pl
Abstract. Modern mobile networks provide access to a variety of ser-
vices. On the one hand, mobile network operators (MNOs) are required
to ensure adequate network coverage, on the other hand, the provided
services should be performed at an appropriate level. Introducing the
assessment and management system quality of service (QoS) is designed
to fulfill the second task. In this paper, we present a correlation analysis
of QoS metrics and radio signal parameters for a web browsing sce-
nario. This research is based on the measurement campaign carried out
in the vicinity of Warsaw, the Polish capital. The measurements were
carried out for two MNOs by a professional company that deals with
QoS assessment in mobile networks. One of the MNOs provides services
using Universal Mobile Telecommunications Service (UMTS) and Long-
Term Evolution (LTE) technologies, while another is based only on LTE
technology. For calculations, we used the Pearson correlation coefficient
and linear regression. The obtained results indicate which of the analyzed
metrics are characterized by a strong, medium, and weak correlation.
Keywords: Quality of Service (QoS) · Quality of Experience (QoE) ·
Mobile network · Drive test · Measurements · Correlation analysis ·
Pearson correlation coefficient · Web browsing
1 Introduction
The quality of services in mobile networks is currently of intense importance both
for telecommunications infrastructure operators, their end-users (subscribers and sub-
scribers of the services), and regulators of telecommunications markets. For the mobile
network operator (MNO), knowledge of quality of service (QoS) and experience (QoE)
is a valuable tool for assessing the network and the level of provided services but is
also used to determine the causes of network failures, unavailability of services, and
performance fluctuations. In addition, MNOs can use the conclusions from the statisti-
cal analysis of QoS and QoE measurements to assess the performance of devices and
xxxiv D. Zmysłowski and J. M. Kelner
networks offered by suppliers at the stage of proof of concept (PoC) pilot studies of new
network solutions.
QoS indicators are getting you to research, analyze and evaluate the technical aspects
of the functioning of the network providing services in terms of meeting the requirements
that initially had been set. By analyzing the values of QoE key performance indicators
(KPIs), one can characterize the service state in a given network from the user’s per-
spective. Using the possibility to assess QoS and QoE parameters is also essential for
telecommunications market regulators. They can indicate to end-users the current service
state offered by individual networks and compare them from the user’s perspective. The
market regulators may also conduct studies on how MNOs are in line with declarations
submitted in auction and concession procedures of network and service development,
their scope, range, and quality [1, 2].
In this paper, we evaluate the correlation between radio signal parameters and QoS
metrics. Such an assessment may be the basis for modeling one of the parameters based on
the other [3]. The authors of [4] propose an innovative approach to service selection that
not only considers QoS correlations of services but also accounts for QoS correlations of
user requirements. On the other hand, the correlational analysis of QoS metrics allows
for the selection of uncorrelated ones for the assessment of MNOs in the broader aspect
of the provided QoS and QoE [2].
2 Problem Solution
Correlation is one of the fundamental tools used in the analysis and processing of data and
signals. The correlation analysis allows you to find similarities and relationships between
two variables, properties, features, signals, or processes. The correlation methods are
also used in QoS research and methods that can approximate service composition. The
purpose of using correlation analyses is to determine the optimal path of service delivery
[3].
The main concern of our work was to research the correlation between parame-
ters of radio signals of mobile networks such as reference signal received power (RSRP)
and signal-to-interference-plus-noise ratio (SINR) for Long-Term Evolution (LTE) tech-
nology, reference signal code power (RSCP) and downlink carrier-to-interference ratio
(EC/IO) for Universal Mobile Telecommunications Service (UMTS) technology, aver-
age throughput (Th), throughput for first second of transmission (Th1s), and different
time metrics of QoS, e.g., time (TD) and session duration (SC), first round trip time
(RTT1), average round trip time for rest transmission (RTT2), time to first byte (T1B),
time to 50% of volume (T50V), time to first 500 kB of transmission (T500kB).
The basis of the performed correlation analysis are measurements made by a profes-
sional company Systemics-PAB Group for the Polish regulator of the telecommunica-
tions market, i.e., the Office of Electronic Communications (UKE). The measurements
were carried out near Warsaw within one week time for mobile networks of two MNOs
(i.e., MNO#1 and MNO#2) that operating in particular in the third (3G) and fourth
generation (4G) standards, i.e. UMTS and LTE, respectively [3, 5].
Relationships Between QoS/QoE Metrics xxxv
The measurements were performed using a professional test-bed by Rohde &
Schwarz, which consisted of SwissQual Diversity Smart Benchmarker Rel. 20.3 with
SwissQual QualiPoc software installed in Samsung Galaxy S20 +5G (SM-G986BDS)
terminals. These terminals support the carrier aggregation technology and all bands used
by MNOs in Poland. A passive scanner, Rohde & Schwarz TSME6, was used to eval-
uate the quality and strength of reference signals in the UMTS and LTE networks. The
scanner supports all frequency bands used in cellular networks [3].
In this paper, we focus on the issue of QoS assessment in mobile networks. Based on
the measurements made during drive tests, the relationships between QoS metrics and
parameters that define the received signal power and quality in mobile networks, i.e.,
RSRP and SINR or RSCP and EC/IO, are analyzed. The analysis was carried out for
selected measurement scenarios of browsing websites. In our studies, we use the Pearson
correlation coefficient (PCC) and linear regression between parameters. The determined
PCCs for the parameter pairs indicated a relationship between the signal parameters
(i.e., radio signal power and quality) or the QoS metrics like throughput and time data
transmission ones. Deviations and regression lines for the metrics’ pairs with significant
PCCs were determined. Exemplary results of correlation analysis for MNO#1 and LTE
technology, we show in the following Table 1.
Table 1. PCCs for MNO#1 and LTE technology.
3 Summary
The carried-out correlation analysis of the data recorded during the test drive shows that
the mentioned QoS metrics and signal parameters are related. The strong correlation we
may see between time metrics (e.g., TD-SD, TD-T50V, T1B-T500kB), throughput (i.e.,
Th-Th1s) or signal parameters (e.g., SINR-RSRP for LTE). Between some QoS metrics
and radio signal parameters, the correlation is from medium to weak. On the one hand, a
better quality of the channel and received signal translates into a higher level of provided
QoS, which is an obvious conclusion. However, the relations between these metrics
xxxvi D. Zmysłowski and J. M. Kelner
are not trivial and are characterized by a certain spread. Its determination is possible
by assessing the linear regression between these parameters. For small absolute PCC
values, the relationship between the metrics is usually nonlinear. On the other hand, the
Shannon formula shows that capacity, (and indirectly also throughput) is closely related
to signal quality (i.e., SINR) and bandwidth. Therefore, the radio resources enabled
by MNO for users to realize the service should additionally consider in the correlation
analysis between QoS and signal parameters. Our future studies will be focused on the
search for non-linear relationships between the analyzed parameters.
Only the data sets in which all the discussed parameters were determined were used
in the correlation analysis. The analysis of incomplete data that was not considered also
showed that the amount of rejected data was smaller in one of the MNOs that provided
services only via the LTE network compared to the other, which used switching between
UMTS and LTE technologies. This proves that the resignation from UMTS technology
in favor of using its radio resources by LTE is more effective from the viewpoint of the
MNOs and users.
The obtained results allow, on the one hand, to model changes in QoS metrics based
on the values of signal parameters (e.g., RSRP or SINR). On the other hand, the con-
ducted analysis allows limiting the number of correlated parameters considered in the
assessment of QoS-related phenomena or processes.
Acknowledgements. This work was financed by the Military University of Technol-
ogy under Research Project no. UGB/22-740/2022/WAT on “Modern technologies of
wireless communication and emitter localization in various system applications”.
The authors would like to thank the President of UKE, Dr. Jacek Oko, for providing
the measurement data made by the Systemics-PAB Group company, which implemented
QoS assessment campaigns in mobile networks on behalf of UKE.
References
1. Mongay Batalla, J., Sujecki, S., Oko, J., Kelner, J.M.: Cost-effective measurements
of 5G radio resources allocation for telecom market regulator’s monitoring. In: 2022
19th ACM International Symposium on Performance Evaluation of Wireless Ad Hoc,
Sensor and Ubiquitous Networks (PE-WASUN), pp. 83–90. ACM, New York (2022)
2. Zmysłowski, D., Kelner, J.M.: The concept of application of the Wroclaw taxon-
omy for QoS assessment in mobile networks. In: Decker, S., Domínguez Mayo, F.J.,
Marchiori, M., Filipe, J. (eds.) 2022 18th International Conference on Web Infor-
mation Systems and Technologies (WEBIST), pp. 485–494. SCITEPRESS, Setúbal,
Portugal (2022).
3. Zmysłowski, D., Kelner, J.M.: Drive test-based correlation assessment of QoS param-
eters for exemplary measurements scenario in suburban environment. In: Decker, S.,
Domínguez Mayo, F.J., Marchiori, M., Filipe, J. (eds.) 2022 18th International Con-
ference on Web Information Systems and Technologies (WEBIST), pp. 485–494.
SCITEPRESS, Setúbal, Portugal (2022).
Relationships Between QoS/QoE Metrics xxxvii
4. Li, D., Ye, D., Gao, N., Wang, S.: Service selection with QoS correlations in distributed
service-based systems. IEEE Access 7, 88718–88732 (2019)
5. Zmysłowski, D., Kelner, J.M., Falkowski-Gilski P.: Mobile networks’ analysis in
terms of QoS performance assessment. In: 2022 19th EAI International Confer-
ence on Mobile and Ubiquitous Systems: Computing, Networking and Services (EAI
MobiQuitous). Springer, Heidelberg (2022)
