The document summarizes an experiment conducted to investigate the performance of LTE systems for high-speed trains carrying massive numbers of onboard users. The experiment tested handover latency and throughput reduction as the number of users increased from 16 to 120. Results showed that handover latency increased exponentially, reaching 1 second for 120 users traveling at 300km/h. Throughput also decreased with higher user numbers. The experimental setup isolated Doppler effects to specifically measure the impact of frequent handovers due to high speed travel.

1. Confidential – Version 1.6
Load-Stress Test of Massive Handovers for LTE System in High Speed Trains
Ali Parichehreh*, Umberto Spagnolini†, Paolo Marini*, Alberto Fontana*, Paolo Timelli*
*Prisma Telecom Testing, †Politecnico di Milano, Milano, Italy
{alip,paolom,albertof,paolot}@prismatelecomtesting.it, {umberto.spagnolini}@polimi.it
Objectives
Creating an in-lab experiment to investigate the
performance of commercial eNBs serving multiple cells
arranged with a linear topology coincident with train-track
network plan with massive on-board UEs.
Characterizing the impairment effects of massive load on
the end users QoS:
Handover induced latency
Throughput reduction
Finding an upper bound of LTE system performance for
the QoS of N={16, 32, 45, 72, 88, 120} users by isolating
Doppler effect.
Scenario and Methodology
Results
Results
Conclusions
Exponential functions can be fitted to the HO-induced
latency with scaling parameter depending on the number of
UEs.
HO-induced latency increases up to 1s for 120 on-board
UEs at the speed of 300km/h.
Experimental results are purposely without Doppler effect
on each UE to isolate the impact of HO on the QoS of
onboard users vs the number of UEs.
The results here are the reference upper-bound for the UEs
density and the specific eNB manufacturer.
LTE Relay LTE Relay
HO region
LTE
servicerate
Moving direction
Uu
Un
Scenario
Method
UE simulation environment
A. Parichehreh, U. Spagnolini, P. Marini, A. Fontana, P. Timelli, Load-Stress Test of Massive Handovers for LTE System in High Speed Trains, European Conference on Network and Communication, Athens, Greece, 2016
AirMosaic: is a Java application which runs on the user
PC (Windows, Linux and Mac supported).
Test Manager: is an application which runs on a dedicated
Linux server, called TMW (TestManager Workstation).
Line Server Unit (LSU): All the Layer 2 protocol
processing and the IP Traffic management is performed by
the LSU Protocol Stack SW which runs on dedicated multi
core server based on a multi OS QNX contumized by
Prisma.
Software Defined Radio (SDR): The LTE Layer 1 digital
signal processing is performed by the SDR processing
cards designed by Prisma: one baseband and one booster
board for load balancing.
Experimental Test-Bed (UeSIM)
Two-tier architecture of HST connectivity. L1/MRN on the roof of the HST.
Train-track with K=200 eNBs, cell-size D=1km, v=300km/h, with
on-board relaying antennas (L1 relay), throughput for each UE
(on the top), and HO (shaded area); UeSIM equipments for a
set of N simulated Ues.
D=1kmMoving direction v=300kmpha)
b)
eNB1 eNB2 eNB3 eNBK
0Throughput
Simulated group of UEs via UeSIM LTE eNB & core network
0 0 0 0group of UEs
PHY
UE N
PHY
MAC
MAC
MAC
MAC
MAC
MAC
UE 1
UE 2
UE N
UE 1
UE 2
PHY level
settings
BB InterfeNodeB
cable
eNodeB
Cell k
Cell k+1
LTE HO procedure
On-board UEs
UE1 –CH2
UE1-CH1
UE2 –CH2
UE2-CH1
UE2 –CH2
UEN-CH1
UE1 (TCP connection)
UE2 (TCP connection)
UEN TCP connection)
Trafficanalysis
(QoS,throughput)
eNB
port 1
port 2
cable
eNB
UE1 (TCP connection)
UE2 (TCP connection)
UEN (TCP connection)
PDCP
RLC
MAC
PDCP
RLC
MAC
PDCP
RLC
MAC
(L1-L3) PHY
Multiusertraffic
generator
(video,HTTP,FTP)
FTP serverPE-Wireshark
Ethernet Uu,Un SGi
(b)
10
0
10
1
10
2
10
3
0
0.2
0.4
0.6
0.8
1
t= RRCcomplete
- RRCconfig
[ms]
ECDF
Experimental 4X4
F(t;t°
)=1-e-0.275t
Experimental 30x4
F(t;t°
)=1-e-0.002t
(a)
0 200 400 600 800 1000 1200
0
0.2
0.4
0.6
0.8
1
t= RRCcomplete
-RRCconfig
[ms]
ECDF
4x4
8x4
15x4
18x4
22x4
30x4
10x4
8x4 15x4 18x4 22x4 30x44x4 10x4
Effect of the massive on-board UEs on the downlink PDSCH
throughput (standard deviations in shaded-gray area). HST
with M=4 carriages and {8, 30} UEs per carriage, at the
speed v=300km/h. UEs in each carriage are connected to the
eNBs through L1 relay node.
Average
Throughput
Average
Throughput
8x4
30x4
Empirical CDF of HO interruption time vs the number of UEs. (a)
Experimental results; (b) Curves fitted (solid lines) to the empirical CDFs
(dashed lines), 𝐹 𝑡, 𝜏° = 1 − 𝑒−0.275𝑡 fitted for 4×4 UEs, and 𝐹 𝑡, 𝜏° = 1 −
𝑒−0.002𝑡 fitted for 30×4 UEs; HST with M=4 carriages, v=300km/h