Page 226 - ITU Kaleidoscope 2016

2016 ITU Kaleidoscope Academic Conference

5. CONCLUSIONS [4] ITU-T, Q.3960: “Framework of Internet speed measurements
for the fixed and mobile networks”. To be approved in 2016.
This paper has evaluated multiple experimentation [5] ITU-T, SG-11. TD 857: “Draft new Recommendation ITU-T
proposals, findings and conclusions as well as the Q.TM_Int_sp_test: Testing methodologies of internet speed
explanation of the precise research momentum of mobile measurement system to be used on the fixed and mobile
Internet performance measurements. We analyze the networks”, September 2015.
experimentation stair with four different approaches and [6] ETSI TS 103 222-1, Part 1, Section 6: “Broadband Speed
present our experiences in relation to QoS measurements Test (BST) – Measurement method of data transmission
and TCP study over LTE. Due to the intrinsic variability of speed and transit time”. November 2015.
[7] Matt Mathis and Al Morton, “Model Based Metrics for Bulk
mobile networks, it is been addressed the need for proper
Transport Capacity,” Internet-Draft draft-ietf-ippm-model-
TCP understanding before the implementation of based-metrics-08,Internet Engineering Task Force, Jul. 2016.
appropriate performance measurements. Four main
[8] S. Shashikant et al., “Survey of Bandwidth Estimation
conclusions are inferred from the described experiments: a) Techniques in Communication Networks,” Wireless Personal
The use of real UEs is mandatory. b) The device could have Communications, vol. 83, no. 2, pp. 1425–1476, 2015.
a great impact for quality reporting reasons. c) The [9] R. Merz et al., “Performance of LTE in a High-velocity
impossibility to perform realistic measurements in emulated Environment: A Measurement Study” AllThingsCellular
testbeds in an isolated way. d) Large-scale deployments are Workshop (SIGCOMM), pp. 47–52, 2014.
needed for a full study and understanding. [10] H. Jiang et al., “Tackling bufferbloat in 3G/4G networks,”
IMC '12 Proceedings of the 2012 ACM conference on
The analysis of short transmissions, to help identify TCP
Internet measurement conference, Pages 329-342, 2012.
performance and measurement constraints in the Slow-Start
[11] S. Alfredsson et al.,“Cross-layer analysis of TCP performance
phase, has shown that Hybrid Slow-Start has a huge barrier
in a 4G system,” in SoftCOM, Sep. 2007, pp. 1–6.
related to its deceleration when a delay peak is detected. [12] J. Huang et al., “An in-depth study of LTE: effect of network
The sensitivity seems to be not adapted to mobile networks protocol and application behavior on performance”,
variability and it has been concluded that the closest to real- SIGCOMM Comput. Commun. Rev. 43, 4, pp. 363-374,2013.
world deployments, the harder impact on performance. It [13] I. Johansson, “Congestion control for 4G and 5G access,”
denotes a strong need for verification and analysis over a Internet Engineering Task Force, Internet-Draft draft-
real-world deployment so as to be able to propose johansson-cc-for-4g-5g-02, Jul. 2016, Work in Progress.
[14] H. Tazaki et al., “Direct Code Execution: Revisiting Library
enhancement to adapt CCAs to mobile networks and OS Architecture for Reproducible Network Experiments,” in
overcome such constraints.
CoNEXT ’13. New York, NY, USA: ACM, 2013.
Finally, through the analysis of TCP’s long-term [15] http://www.fed4fire.eu/testbeds/
performance over LTE, it has been proven that the [16] S. Bouckaert et al., “The w-iLab.t Testbed”, Testbeds and
experimentation stair is capable of confirming macroscopic Research Infrastructures. Development of Networks and
findings and behaviors amongst different steps. Considering Communities: 6th International ICST Conference,
all the aforementioned information, a large-scale mobile TridentCom 2010, Berlin, Germany, May 18-20, 2010.
deployment campaign has been proposed to dispel doubts [17] J. Garcia, S. Alfredsson & A. Brunstrom: Delay metrics and
delay characteristics: A study of four Swedish HSDPA+ and
regarding QoS measurements dichotomy, TCP-based
findings and so on. LTE networks. In EuCNC, 2015.
[18] Anna Brunstrom, “Mobile Broadband Measurement Platform
For future work, we will apply the suggested proposal in
- MONROE”, IETF-93 Proceedings. Prague, 19 July 2015.
the MONROE [18] open testbed to validate the framework [19] V. Bajpai and J Schönwälder, “A Survey on Internet
with sufficient resources in terms of static and mobile nodes Performance Measurement Platforms and Related
and to verify the suitability and feasibility of QoS Standardization Efforts”, IEEE Communication Surveys &
measurement methodologies. It is also important to clarify Tutorials, vol. 17, no. 3, pp. 1313-1341, Third quarter 2015.
that the validation framework has the capability to back-up [20] Matt Mathis, “Why is BTC so hard?”, IP Performance
explained findings related to TCP performance over LTE Metrics Workgroup, IETF-86 Proceedings, March 2013.
[21] Al Morton, “Improved Internet speed tests can enhance QoS
and give a deep input to further developments.
th
and QoE”, 4 International Workshop on Perceptual Quality

of Systems (PQS 2013), 2013.
ACKNOWLEDGEMENTS [22] IETF RFC 7594 “A framework for Large-Scale Measurement

This work has been funded by the Spanish Ministerio de of Broadband Performance (LMAP)”.
Economía y Competitividad (MINECO) under grant [23] E. Atxutegi et al., "Why we still need standardized Internet
speed measurement mechanisms for end users", ITU
TEC2013-46766-R: “QoEverage” and by the European
Kaleidoscope 2015, Barcelona, Spain, 9-11 December 2015.
Union H2020-ICT grant agreement no. 644399
[24] Jose Oscar Fajardo, “Analysis of channel quality reporting
(MONROE).
mechanisms and their impact into 4G and beyond
REFERENCES technologies”, PhD Thesis, December 2015.

[1] Cisco, “Cisco Visual Networking Index: Global Mobile Data [25] S. Ha and I. Rhee, “Taming the elephants: New TCP slow
Traffic Forecast Update, 2015 – 2020,” Cisco, Tech. Rep., start,” Computer Networks, vol. 55, no. 9, pp. 2092–2110,
2016. Jun. 2011.
[2] ITU-T, Y.1543: Measurements in IP networks for inter- [26] E. Atxutegi et al., "TCP behavior in LTE: impact of flow
domain performance assessment, 2007. start-up and mobility", IFIP Wireless and Mobile Networking
[3] Y. FMIPQoS: Framework for Monitoring the QoS of IP Conference - WMNC, 2016.
network services, May 2016.

– 208 –