=Paper=
{{Paper
|id=Vol-2768/paper1
|storemode=property
|title=Delay effect on VoLTE End-to-End Performance
|pdfUrl=https://ceur-ws.org/Vol-2768/p1.pdf
|volume=Vol-2768
|authors=Alessandro Vizzarri
}}
==Delay effect on VoLTE End-to-End Performance==
https://ceur-ws.org/Vol-2768/p1.pdf
Delay Effect on VoLTE End-to-end Performance
Alessandro Vizzarria
a Radiolabs, Laboratori di Radiocomunicazione, Consorzio Università Industria, Rome, Italy
Abstract
Smart city is a new paradigm indicating a well performing city in terms of a efficiency and sustainability perceived in different
sectors (e.g. mobility, environment, health, space living) by citizens. New cellular technologies as Long Term Evolution (LTE)
are crucial for this new approach due to possibility to enable pervasive reception of digital services by users. In order to be
able to use all LTE innovative features, Mobile Network Operators (MNO) need to also guarantee acceptable QoS perceived by
end user. End-to-end approach for QoS is strongly recommended, especially for IP delay sensitive services like VoIP over LTE
(VoLTE). This work presents analysis of delay impact on VoLTE end-to-end performances in multi-user and multi-service
(VoLTE and HTTP web browsing) environment. Different LTE network scenarios are simulated using OPNET modeler 17.5.
A final comparison of simulation results is provided in order to evaluate delay impact on VoLTE end-to-end performance.
Keywords
LTE, VoIP, VoLTE, End-to-end QoS, QoE, LTE network performance, LTE KPIs, OPNET, IP cloud, Smart cities.
1. Introduction UTRAN (E-UTRAN) and a Core Network called Evolved
Packet Core (EPC). LTE system also introduces a direct
Today our cities need to be smarter, especially in terms management of QoS policies based on bearers and QoS
of efficiency and sustainability perceived by citizen in Class Identifier (QCI) in order to guarantee acceptable
different sectors of city, e.g. mobility, environment, servie reception by end user. QoS policies in LTE are
health, space living [1, 2, 3]. ICT technologies rep- mainly focused on available bandwidth, delay, packet
resent the key factor enabling smart city paradigm. loss ratio, data rate, priority.
Among them cellular technologies, especially Long Term Actual scenario of ICT technologies is characterized
Evolution (LTE) which indicates 4th Generation (4G) by a strong convergence of different ICT networks (wire-
standard for wireless systems, gives to citizens the op- less, wireline, cable). LTE is part of this important
portunity to make smarter their living conditions due multi-network, multi-client and multi-service [9, 10]
to a pervasive reception of digital services. environment where several networks, services and users
This is surely possible also thanks to the incredi- are operating.
ble capabilities in terms of computing of modern digi- In order to guarantee acceptable QoS to end user us-
tal systems that allow the efficient implementation of ing IP transport protocol and best effort type of service
communication systems [4, 5, 6]. (ToS), Mobile Network Operator (MNO) need to inte-
In the first quarter of 2014 number of mobile broad- grate LTE native QoS features with other enhanced
band subscribers was characterized by a very strong techniques based on a QoS end-to-end approach. Es-
growth. As mentioned in [7] they will be expected to pecially in case of IP delay sensitive services like VoIP
growth from 6.8 billion in the first quarter of 2014 to over LTE (VoLTE) [11].
9.2 billion by the end of 2019. Penetration of LTE ter-
minals has grown very quickly: until June 2014 a value
of 240 million subscriptions was reached (35 million of 2. Related works
subscription added on the first quarter of 2014).
LTE is the first 3GPP cellular standard full IP-based. Since LTE is full IP-based wireless standard it only en-
It is able to offer to end users a download data rate up ables entire wireless transmission over Packet Switch-
to 100 Mbps and an upload data rate up to 50 Mbps [8]. ing (PS) paths using IP protocol. All major standard-
LTE is also characterized by several innovative fea- ization entities already treated issue of QoS in LTE us-
tures. Architecture is more flexible and interoperable. ing end-to-end approach. In [12] ETSI provides end-
It is composed by radio access interface called Evolved to-end QoS reference architecture for LTE and a de-
scription of relative management functions. If these
ICYRIME 2020: International Conference for Young Researchers in requirements are mandatory to be implemented, all
Informatics, Mathematics, and Engineering, Online, July 09 2020 kind of QoS policies and strategies focused on how
" alessandro.vizzarri@radiolabs.it (A. Vizzarri)
manage user’s traffic flows are left to MNOs. More-
© 2020 Copyright for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0).
over, devices should be equipped with suitable hard-
CEUR
Workshop
CEUR Workshop Proceedings (CEUR-WS.org)
Proceedings
http://ceur-ws.org
ISSN 1613-0073 ware in order to perform optimal visualization [13].
Figure 1: LTE network topologies used for simulation
In real scenarios many network impairments can with higher priority. Strategies of priority service allo-
occur (e.g. excessive delay caused by network conges- cating are left to operators [20]. This is also confirmed
tion or faults) and can bring a negative effect for ac- by Medbo et al. in [21]. In case of LTE mixed traf-
ceptable service transmission/reception [14, 15]. Op- fic (based on VoIP as first service and real time video
erators need to adopt opportune network management or web browsing as second service), service differen-
policies in order to guarantee a good service percep- tiation and prioritization of delay-sensitive traffic (e.g.
tion by end user, especially in case of full IP-based VoIP) can improve its performance without affecting
network as LTE. In scientific literature several tech- performances of other delay-insensitive services. Thus
nical papers are focus on it. Vizzarri et al. in [16] is due to small size of VoIP packets.
presents a review of most important papers focused
on end-to-end QoS approach in LTE networks: first
studies are only concentrated in E-UTRAN interface, 3. VoLTE end-to-end QoS
the last ones consider entire transmission chain (both Assessment
E-UTRAN and EPC).
Network management techniques have also a cru- Since an efficient end-to-end approach to quality of
cial role for QoS guaranteeing procedures. Horvath service needs to analyze both quality of content deliv-
in [17] presents a new LTE QoS Signaling (LQSIG) pro- ered (voice quality in case of VoLTE) and network per-
tocol able to guarantee a resource reservation for data formances, a QoS assessment based on the most im-
path transmission compliant with LTE QCI target val- portant Key Performance Indicators (KPIs) is needed [22].
ues. S. Shen et al. in [18] propose a new LTE per- In this work following KPIs are considered in case of
formance management framework based on CoS/QoS VoLTE [23]:
mapping table: main scope is to identify relationship
among QoS informations in LTE and operator trans- • Mean Opinion Score (MOS)
port network. In particular LTE service QCIs are com-
• End-to-end delay
pared with QoS information in case of IP protocol (DSCP
value), carrier Ethernet (802.1p value) and MPLS (EXP • Packet loss
value). Margoc et al. in [19] analyze QoS in LTE sys-
tems that confirms better performances for services • Jitter
2
Table 1
KPIs analyzed for LTE services.
Figure 2: End-to-end VoLTE scenario
MOS is a scalar term included in range 1-5. It indi-
cates quality of a voice, VoIP and VoLTE call. Packet
loss gives a percentage indication of number of packets
lost during transmission. Jitter is variance of packet
inter-arrival time. End-to-end delay can be summa-
rized by following formula:
𝐷𝑒2𝑒−𝑡𝑜𝑡 = 𝐷1 + 𝐷2 + 𝐷3 (1) Table 2
simulated scenarios configuration.
where
𝐷1 = 𝐷𝐸𝑛𝑐𝑜𝑑𝑖𝑛𝑔 + 𝐷𝑃𝑟𝑜𝑐𝑒𝑠𝑠 + 𝐷𝑝𝑎𝑐𝑘𝑒𝑡𝑖𝑧𝑎𝑡𝑖𝑜𝑛
𝐷2 = 𝐷𝑇 𝑟𝑎𝑛𝑠𝑚𝑖𝑠𝑠𝑖𝑜𝑛 + 𝐷𝑄𝑢𝑒𝑖𝑛𝑔 + 𝐷𝑃𝑟𝑜𝑝𝑎𝑔𝑎𝑡𝑖𝑜𝑛 + 𝐷𝐵𝑎𝑐𝑘𝑏𝑜𝑛𝑒
𝐷3 = 𝐷𝐽 𝑖𝑡𝑡𝑒𝑟_𝑏𝑢𝑓 𝑓 𝑒𝑟_𝑑𝑒𝑙𝑎𝑦 + 𝐷𝐷𝑒𝑐𝑜𝑑𝑖𝑛𝑔 + 𝐷𝑃 𝑙𝑎𝑦𝑏𝑎𝑐𝑘
Figure 2 shows a graphic representation.
D1 is delay affecting VoLTE caller (𝑈 𝐸_1). In par-
ticular DEncoding is delay due to voice codec utilized,
DProcess to hardware processing, Dpacketization due
to packetization. D2 is delay affecting entire LTE net-
work. It is composed by DTransmission due to Uplink
and downlink transmission of sender (𝑈 𝐸_1) and re-
ceiver (𝑈 𝐸_2), DQueuing due to queuing and schedul-
ing management of packets, DPropagation due to en-
tire propagation from sender to receiver, DBackbone
due to influence of transmission across MNO back-
bone.
D3 is delay affecting VoLTE callee (𝑈 𝐸_2). In par- 4.1. Scenarios
ticular 𝐷𝐽 𝑖𝑡𝑡𝑒𝑟_𝑏𝑢𝑓 𝑓 𝑒𝑟_𝑑𝑒𝑙𝑎𝑦 is delay due to buffering ca-
Four different scenarios are simulated changing LTE
pacity, 𝐷𝐷𝑒𝑛𝑐𝑜𝑑𝑖𝑛𝑔 due to decoding phase, 𝐷𝑃 𝑙𝑎𝑦𝑏𝑎𝑐𝑘 network topology (with or without IP cloud) and traf-
due to terminal software reproducing decoded audio. fic flows (single or mixed).
Main purpose of this work is to analyze influence of Scenario n. 1 is characterized by two UEs: 𝑈 𝐸_1
additional impairments on VoLTE KPIs. Impairments (caller) is performing a VoLTE call to 𝑈 𝐸_2 (callee) us-
are represented by mixed traffic (VoLTE and HTTP ing a direct link. Traffic flow is single.
web browsing) and delay introduced by presence of IP Scenario n. 2 is characterized by three UEs and one
cloud [24]. D2 is main involved factor. This situationHTTP web server: 𝑈 𝐸_1 and 𝑈 𝐸_2 are performing a
is typical in case of interaction between LTE network VoLTE call while 𝑈 𝐸_3 is requesting a web service ses-
and backbone network. sion to HTTP server. Traffic flow is mixed: VoLTE and
Table 1 shows most KPIs measured in case of VoLTE HTTP web browsing services are performed by UEs.
and HTTP services [25]. Scenario n. 3 is similar to scenario n. 1: 𝑈 𝐸_1 is
perfoming a VoLTE call to 𝑈 𝐸_2, but link among them
is interrupted by IP cloud, responsible for addition of
4. Simulations packet discard ratio (1%) and delay (01. seconds) to
Simulation activities of VoLTE services are performed entire transmission chain.
using OPNET Modeler 17.5 PL6 software tool. Scenario n. 4 is similar to scenario n. 2: 𝑈 𝐸_1 and
3
• LTE_wkstn: LTE workstation or UE.
• UE_1: VoLTE source (or caller)
• UE_2: VoLTE destination (callee)
• UE_3: requesting HTTP web session
• lte_enodeb_3sector_slip4_adv_1_upgvrade:
(a) LTE e-NodeB with 3 sectors. Two different e-
NodeBs are considered: eNB_1 serving UE_1 and
UE_3, eNB_2 serving UE_2
• lte_epc_atm8_ethernet8_slip8_adv: LTE EPC
node
• Ethernet_server: HTTP server
• PPP_DS3: link model for LTE nodes
(b)
• 100baseT: link model for HTTP web server
OPNET modeler management nodes are:
• app_config: application configuration node
• profile_config: profile configuration node
• lte_attr_definer_adv: LTE attribute definer node
(c) 4.2.2. LTE settings
For all simulated scenarios entire LTE network is mod-
eled using parameters listed in Table 3.
4.2.3. Application configuration
In this paper voice and HTTP applications are selected
among those available in OPNET modeler. New voice
(d) application created is named VoLTE. It is launched with
Figure 3: simulated scenarios. a start offset of 20 seconds till the end of simulation
period. As requested by 3GPP LTE standard, the same
VoLTE application is carried out over EPS bearer with
QCI 1 (GBR) and ARP 1.
𝑈 𝐸_2 are performing a VoLTE call using a direct link
Table 4 lists the main characteristics of VoLTE ap-
interrupted by IP cloud, 𝑈 𝐸_3 is performing HTTP
plication.
web session.
New HTTP application is named HTTP. It is launched
Table 2 lists the most important features of each
with a start offset of 40 seconds and it is characterized
simulated scenario.
by:
4.2. OPNET Settings • N. 1 web page with dimension: 1000 Bytes (con-
stant distribution)
4.2.1. Network Topology
• N. 5 medium images with dimension: uniformly
Figure 3 shows different LTE network topologies used variable from 500 to 2000 bytes
during simulation activity.
Figure 3a is used in scenario 1, 3b in scenario 2, 3c • N. 2 short videos with dimension: uniformly vari-
in scenario 3, 3d in scenario 4. able from 10000 to 350000 bytes
Simulation area is a typical campus area 10 × 10 Km. Table 5 lists the main characteristics of HTTP appli-
Models of LTE network nodes are: cation.
4
Table 3 Table 5
LTE Network settings HTTP Application features.
Table 6
Voice profile settings.
Table 4
VoLTE Application features.
application, the second one is related to configuration
of VoLTE profile.
It means that VoLTE packets are going to be sent af-
ter 40 seconds from simulation starting. HTTP profile
4.2.4. Profile configuration uses a start offset of 40 seconds. Since HTTP appli-
cation has also a start offset of 40 seconds, IP pack-
In order to simulate VoLTE and HTTP services, two ets of HTTP session are going to be delivered after 80
different OPNET profiles are created: VoLTE Profile seconds after simulation start. Simulation period is 3
and HTTP profile. Main settings of profiles are listed minutes for all scenarios.
in Table 6 and Table 7.
VoLTE Profile uses a unique application, VoLTE ap-
plication. Since Voice Profile uses also a start offset of 4.3. Simulation results
20 seconds, there are two different start offset of 20 sec- In this paragraph main simulation results based on KPIs
onds: the first one is related to configuration of VoLTE listed in Table 1 are presented. Figure 4-Figure 8 show
5
Table 7
HTTP profile settings.
Figure 5: Simulation results: end-to-end packet delay.
Figure 4: Simulation results: MOS mean value.
Figure 6: Simulation results: VoLTE traffic sent.
graphic representation of each scenario in terms of MOS
mean value, end-to-end packet delay, VoLTE traffic sent, get values fixed by QCIs. However MOS mean value is
VoLTE traffic received, HTTP Page response Time. far from target value (see table IV) typical of GSM EFR
Table 8 shows comparison of final simulation results voice codec. Scenarios 3 and 4 are affected by higher
for each scenario. value of end-to-end delay and packet loss than previ-
A general comparison among all scenarios under- ous scenarios. Worst performance of scenario n. 4 is
lines better KPIs performance in scenarios 1, 2 (with- also evidenced. Combination of mixed traffic together
out IP cloud) than ones measured in scenarios 3, 4 (with with impairments caused by IP Cloud (1% packet dis-
IP cloud). Simulation results of scenarios 1 and 2 indi- card ratio and 0.1 second delay) determines KPI end-
cate a general behavior of KPIs quite near LTE QoS tar- to-end values around 1.7 in terms of MOS, around 31%
6
Table 8
Service performance results.
Figure 7: Simulation results: VoLTE traffic received. VoLTE in a multi-user and multi-service environment.
Four different scenarios are simulated using OPNET
Modeler software tool. Simulation results show a gen-
eral worsening of all measured KPIs either when HTTP
web browsing is added to VoLTE either network im-
pairments due to IP cloud are introduced. Scenario
characterized by both of them has the worst perfor-
mance.
Network impairments due to IP cloud are the heav-
ier factors for a general decrease of VoLTE end-to-end
performance. Improvements in LTE network trans-
mission are necessary both in user and control plane
in order to improve end-to-service quality perceived
by end user and to reach KPI values compliant with
requirements fixed by QCIs. Future works are going to
investigate network improvements in entire LTE trans-
mission chain and additional techniques able to en-
hance QoS management (e.g. queue management, user
priority).
Figure 8: Simulation results: HTTP Page response Time.
References
[1] G. Lo Sciuto, G. Capizzi, S. Coco, R. Shikler, Ge-
in terms of packet loss ratio and 01 seconds in terms
ometric shape optimization of organic solar cells
of packet delay. These negative performances justify
for efficiency enhancement by neural networks,
adoption of strong network management techniques
Lecture Notes in Mechanical Engineering (2017)
by MNO in order to reach maximum value fixed by
789–796.
LTE QCI (10%) using end-to-end QoS approach.
[2] G. Capizzi, G. Lo Sciuto, G. Cammarata, M. Cam-
marata, Thermal transients simulations of a
5. Conclusion building by a dynamic model based on thermal-
electrical analogy: Evaluation and implementa-
VoLTE is a crucial service for LTE networks. Purpose tion issue, Applied Energy 199 (2017) 323–334.
of this work is to evaluate end-to-end performance of [3] C. Napoli, G. Pappalardo, E. Tramontana, An
7
agent-driven semantical identifier using radial ios, Future Internet 9 (2017) 71.
basis neural networks and reinforcement learn- [15] C. Napoli, F. Bonanno, G. Capizzi, Exploiting
ing, in: Proceedings of the XV Workshop ”Dagli solar wind time series correlation with magne-
Oggetti agli Agenti”, volume 1260, CEUR-WS, tospheric response by using an hybrid neuro-
2014. wavelet approach, Proceedings of the Interna-
[4] G. Cardarilli, L. Di Nunzio, R. Fazzolari, M. Re, tional Astronomical Union 6 (2010) 156–158.
Fine-grain reconfigurable functional unit for em- [16] S. Forconi, A. Vizzarri, Review of studies on end-
bedded processors, in: Conference Record - to-end qos in lte networks, in: AEIT Annual Con-
Asilomar Conference on Signals, Systems and ference 2013, IEEE, 2013, pp. 1–6.
Computers, 2011, pp. 488–492. doi:10.1109/ [17] G. Horváth, End-to-end qos management across
ACSSC.2011.6190048. lte networks, in: 2013 21st International
[5] G. Capizzi, S. Coco, G. Lo Sciuto, C. Napoli, A Conference on Software, Telecommunications
new iterative fir filter design approach using a and Computer Networks-(SoftCOM 2013), IEEE,
gaussian approximation, IEEE Signal Processing 2013, pp. 1–6.
Letters 25 (2018) 1615–1619. [18] L. Li, S. Shen, End-to-end qos performance man-
[6] C. Napoli, G. Pappalardo, E. Tramontana, A agement across lte networks, in: 2011 13th
hybrid neuro–wavelet predictor for qos con- Asia-Pacific Network Operations and Manage-
trol and stability, in: AI*IA 2013: Advances ment Symposium, IEEE, 2011, pp. 1–4.
in Artificial Intelligence, volume 8249 of Lec- [19] A. Markoč, G. Šišul, Quality of service in mobile
ture Notes in Computer Science, Springer Inter- networks, in: Elmar International Symposium
national Publishing, 2013, pp. 527–538. doi:10. Electronics in Marine, 2013, pp. 263–267.
1007/978-3-319-03524-645. [20] I. Benedetti, R. Giuliano, C. Lodovisi,
[7] Ericsson, Ericsson Mobility Report, Technical Re- F. Mazzenga, 5g wireless dense access net-
port, 2014. work for automotive applications: Opportunities
[8] 3GPP, Overview of 3GPP Release 8: Summary of and costs, in: 2017 International Conference
all Release 8 Features. Release 8 V0.0.3, Technical of Electrical and Electronic Technologies for
Report, November 2008. Automotive, IEEE, 2017, pp. 1–6.
[9] T. Magedanz, Towards smart city service deliv- [21] J. Medbo, I. Siomina, A. Kangas, J. Furuskog,
ery and control platforms, National Research Propagation channel impact on lte positioning
Council of Thailand (NRCT) Annual Meeting, accuracy: A study based on real measurements
Bangkok ( Thailand) (August 26, 2013). of observed time difference of arrival, in: 2009
[10] M. Carminati, O. Kanoun, S. L. Ullo, S. Mar- IEEE 20th International Symposium on Personal,
cuccio, Prospects of distributed wireless sensor Indoor and Mobile Radio Communications, IEEE,
networks for urban environmental monitoring, 2009, pp. 2213–2217.
IEEE Aerospace and Electronic Systems Maga- [22] T. Gorman, H. Larijani, A. U. H. Qureshi, Voice
zine 34 (2019) 44–52. over lte quality evaluation using convolutional
[11] A. Vizzarri, Analysis of volte end-to-end quality neural networks, in: International Joint Confer-
of service using opnet, in: 2014 European Mod- ence on Neural Networks (IJCNN), 2020, pp. 1–7.
elling Symposium, IEEE, 2014, pp. 452–457. [23] M. Voznak, M. Halas, F. Rezac, L. Kapicak, Delay
[12] ETSI, Digital cellular telecommunications sys- variation model for rtp flows in network with pri-
tem (Phase 2+); Universal Mobile Telecommuni- ority queueing, in: NEHIPISIC’11, IEEE, Febru-
cations System (UMTS); LTE; End-to-end Qual- ary 2011, p. 344–349.
ity of Service (QoS) concept and architecture. TS [24] M. Di Mauro, A. Liotta, Statistical assessment
123.207, Technical Report, Jun 2010. of ip multimedia subsystem in a softwarized
[13] G. Cardarilli, L. Di Nunzio, R. Fazzolari, D. Gia- environment: A queueing networks approach,
rdino, M. Matta, M. Re, L. Iess, F. Cialfi, G. De An- IEEE Transactions on Network and Service Man-
gelis, D. Gelfusa, A. Pulcinelli, L. Simone, Hard- agement 16 (2019). doi:10.1109/TNSM.2019.
ware prototyping and validation of a w-Δdor dig- 2943776.
ital signal processor, Applied Sciences (Switzer- [25] J. Puttonen, T. Henttonen, N. Kolehmainen,
land) 9 (2019). doi:10.3390/app9142909. K. Aschan, M. Moisio, P. Kela, Voice-over-ip per-
[14] F. Mazzenga, R. Giuliano, F. Vatalaro, Fttc-based formance in utra long term evolution downlink,
fronthaul for 5g dense/ultra-dense access net- in: VTC Spring 2008-IEEE Vehicular Technology
work: Performance and costs in realistic scenar- Conference, IEEE, 2008, pp. 2502–2506.
8