Page 219 - ITU Kaleidoscope 2016

ASSESSING INTERNET PERFORMANCE OVER MOBILE
NETWORKS: FROM THEORY TO PRACTICE

Eneko Atxutegi, Jose Oscar Fajardo, Eva Ibarrola and Fidel Liberal

Faculty of Engineering of Bilbao, University of the Basque Country UPV/EHU, Spain
{eneko.atxutegi, joseoscar.fajardo, eva.ibarrola, fidel.liberal}@ehu.eus

ABSTRACT quality and hence its transmission rate. In this regard and
related to auto-inflicted interference, user equipment’s
The proper execution of performance tests is of utmost
(UE’s) mobility pattern and the associated speed severely
importance for the analysis and evaluation of Internet-
impact on Signal to Interference plus Noise Ratio (SINR)
related technologies, protocols and deployment strategies.
and therefore on reported channel quality indicator (CQI),
There are plenty of tools available for experimenters
the available capacity variability and eventually the
ranging from simulation tools, emulation equipment, to
application-level performance [9].
small and large scale experimentation testbeds. Each of
these performance evaluation frameworks introduces a Our work presents an experimentation methodology as a
series of capabilities and drawbacks. Additionally, there is research guidance that could be applied to QoS
a need for using a common methodology to perform measurements or other studies regarding mobile networks,
different performance tests and to create comparable transport protocols and so on. The paper is not intended to
outcomes. Seeking for the standardization of reliable and provide the reader with a novel approach but it will define
comparable Internet speed measurements and trying to and classify the current measurement approaches,
provide end-users with trustworthy measurement tools, presenting them together with their strengths and
different SDOs have standardized different measurement weaknesses, involving distinct environments, testbeds and
methodologies. This paper focuses on the applicability of deployments.
the distinct measurement frameworks and methodologies in The highly variant environment of radio networks is a
the field of mobile Internet, with special focus on the impact handicap for the performance of measurements in its goal to
of using TCP over mobile broadband connections. The achieve the maximum available bandwidth, especially TCP-
paper describes a series of experiments over different based ones. In addition to this, the existence of different
frameworks and, based on the obtained results, identifies TCP implementations (congestion control algorithms –
the implications of the different types of performance tests CCAs-) implies different behaviors that produce different
into the evaluation outcomes and states the need for large outcomes that can be seen as QoS/QoE degradation at
scale measurements. higher layers [10].
Keywords— mobile networks, performance measurements, This paper also analyzes the impact factor that LTE
simulation, experimentation. actors such as UEs and eNodeBs have on the fulfillment of
the proposed measurement approaches. In the same way, it
1. INTRODUCTION
also covers the study of precise TCP-based performance
Mobile data traffic has grown 400-million-fold over the features amongst the steps that the experimentation stair
past 15 years and it is expected to grow nearly eightfold offers so as to analyze the impact that may well have.
between 2015 and 2020 [1]. All these predictions highlight The topic of TCP over LTE has been deeply studied
the importance of mobile networks. In this respect, end- with findings regarding the relation between retransmissions
users’ experienced service (QoE) and fair and contract- and throughput decay [11], undesired Slow-Start impact
compliant Quality of Service (QoS) are two key factors to [12], spectral efficiency and resource assignment under
consider when analyzing the suitability and adaptability of different speeds [9] and promising study and definition of
Mobile Operator service towards end-users. the adaptability and suitability of congestion controls to
TCP-based and model-based quality of service (QoS) mobile networks [13] among other. However, none of the
measurement methodologies has congregated a fair amount abovementioned works have studied various CCAs and the
of researchers, developers and different stakeholders (i.e. impact that the mobility pattern and related speed have at
ITU [2-5], ETSI [6] and IETF’s [7]) reaching no agreement the same time. Our work not only covers such missing
yet on the best feasibility vs. reliability trade-off to assess research areas but also extends some of the mentioned
the key performance indicators (KPIs), leading to a research works. As a validation step for the accuracy of different
and development dichotomy with measurement studies Internet quality measurement methodologies over LTE and
regarding either transport layer [8] or IP layer [7]. future TCP enhancements, a large-scale measurement
Mobile networks, due to their intrinsic variability, campaign proposal is explained with different contexts of
propose a more challenging field to perform QoS and QoE use, experimentation plan and considerations wrap-up.
measurements. Such variability, inflicted by propagation, Summarizing, the main contributions of this paper are:
fading and interference effects impact on the channel

978-92-61-20431-0/CFP1668P-ART © 2016 ITU – 201 – Kaleidoscope