Page 220 - ITU Kaleidoscope 2016

2016 ITU Kaleidoscope Academic Conference

• Experimentation methodology for QoS measurements, are very little (for instance a single computer) and therefore
study of transport protocols and mobile networks. it is an available experimentation way for almost everyone.
• Study of LTE actors’ impact regarding the use of real In the framework of this work, we deployed a hybrid
UEs, usage of realistic environments in terms of fading simulation/emulation setup based on the ns-3 network
simulator. Even though ns-3 allows performing TCP-based
and propagation, distinct devices differences and so on.
measurements with its natively modeled TCP flavors, the
• An analysis regarding TCP performance features.
behavior of these TCP servers differ from real-world
• A proposal of large-scale measurement campaign
servers. Thus, we used Direct Code Execution (DCE)
considering previous conclusions and findings. instead, allowing the execution of a full Linux network
The paper is organized as follows: Section 2 is stack [14] and enabling the tweak of nowadays TCP/IP
responsible for the explanation of the experimentation stack. The target was the understanding of TCP’s behavior
methodology with general related pros and cons and the over LTE under different mobility patterns and speeds. For
standardization efforts regarding QoS measurements and the correct usage and dissemination of LTE, we employed
our linkage with them. Section 3 focuses on the analysis of the LTE/ Evolved Packer Core (EPC) module developed in
TCP performance tests over mobile networks among the project called LENA.
different steps of the experimentation stair. Section 4 Regarding the strengths, two are the most important, the
explains the abovementioned large-scale measurement repeatability among different simulations and therefore the
campaign proposal. Finally, Section 5 covers the ability to fully compare them, and the huge possibility to
conclusions and considerations report. gather parameters and have them under control. In contrast,
applied fadings are synthetic and far from realism, the UEs
2. EXPERIMENTATION METHODOLOGIES are not real and so is the reporting mechanism and finally,

trying to model a real deployment becomes a very tough
This section covers the explanation of the
experimentation methodology divided in two major task with many variables involved.

subsections. Firstly, the experimentation stair will be
2.1.2. Emulated testbed
described (illustrated in Figure 1). In this regard, four
different experimentation approaches are cut off The next approach in experimentation stair comprises
highlighting the strengths and weaknesses of them in terms emulated environments. Precisely in LTE, the so-called
of scalability and suitability to model real deployments. digital radio tests or LTE emulators. These kinds of
Secondly, we review the efforts performed in different emulators play the role of the eNodeB, because they create
standardization bodies concerning mobile Internet the LTE radio signal and all the necessary LTE protocol
performance measurement methodologies and frameworks events to support the attachment and registration of any
and the linkage with our work. LTE device through a radiofrequency cable or over the air.
In our work, we setup an emulated testbed in our lab

with the following machines:
• UEs: Samsung Galaxy S III mobile phone and Samsung
GT-B3730 LTE dongle. Both devices were connected
to the emulator through a radiofrequency cable in order
to avoid undesired signal behavior and for the govern
prohibition in terms of non-registered radiations.
• Server: Linux server with http daemon running to serve
the requested files to the UEs during the experiments.
• Digital radio test set: Aeroflex 7100 is a full Evolved
UTRAN (EUTRAN)/EPC testbed with plenty of tests,
configurable parameters and logging capabilities.

Figure 1. Experimentation stair • Controller: Automation machine for the correct
synchronization and commanding of the involved parts.

The emulated testbed aimed at involving both
2.1. Experimentation stair confirmations of previous findings and the possibility of

The description will follow the stepped structure shown accomplishing studies that were not suitable to achieve
in Figure 1 by the explanation of each experimentation level through raw simulated schemes. In this regard, the emulated
and the deployment associated to them. Table 1 shows a testbed enables using real UEs and therefore managing a
wrap-up regarding the explained pros and cons of each representative reporting to the eNodeB. Besides, it gives the
experimentation level. opportunity to configure LTE parameters straightforward
comparing with simulated environment. Finally, it has the
2.1.1. Simulated environment ability to collect meaningful traces such as CQI, block error
rate (BLER) and so on, making the research task easier to

One of the most used deployments in research is the this respect. On the other hand, the utilized fadings have a
simulated environment. The resources that have to be spent
random feature to make them closer to realism, but required

– 202 –