doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Multi-Layer LTE / Wi-Fi Access Network Selection - results from the SEMAFOUR
project
Date: 2015-07-14
Slide 1
Author: Name Company Address Phone email
Thomas Kürner TU Braunschweig [email protected]
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Outline
• What is the SEMAFOUR project? • SEMAFOUR Vision • Results achieved within SEMAFOUR‘s LTE/WLAN
Traffic Steering use case
Thomas Kürner (TU Braunschweig) Slide 2
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
What is SEMFOUR ?
• SEMAFOUR is a collaborative research project funded by the European Commission within its seventh framework programme
• The goal of SEMAFOUR is to develop a Unified self-management system efficiently operating a heterogeneous mobile network comprising a multitude of radio access technologies and layers
• SEMAFOUR mainly has concentrated on 3GPPP radio technologies, but has considered IEEE 802.11 as well
Thomas Kürner (TU Braunschweig) Slide 3
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SEMAFOUR Key Facts
• Website: www.fp7-semafour.eu • Scheme: EU FP7 STREP (No. 316384) • Duration: 09/2012 – 08/2015 • Effort: 500 Person Months • Budget: 6.1 M€ (total), 3.8 M€ (funding) • Coordinator: Dr. Colin Willcock (Nokia, Germany)
Thomas Kürner (TU Braunschweig) Slide 4
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SEMAFOUR Partners
Thomas Kürner (TU Braunschweig) Slide 5
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SEMAFOUR Vision
Decision Support System
SERVICE PROVIDER
NETWORK OPERATOR
HETEROGENEOUS NETWORK
General Network-Oriented Objectives
Network Status related to General Network-Oriented Objectives
Integrated SON Management
Physical Network
Resources
Physical Network
Resources
Physical Network
Resources
Policy-Based SON Management
Mon
itorin
g &
Di
agno
sis
Operational SON Coordination
Multi-RAT / Multi-layer SON
WLAN Mgmt.
2G / 3G SON
LTE SON
UNIFIED SELF-MANAGEMENT SYSTEM
Thomas Kürner (TU Braunschweig) Slide 6
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Functions for Multi RAT and Multi Layer Networks
Thomas Kürner (TU Braunschweig) Slide 7
time
GSMTCH
LTE3MHz
GSMTCH
GSMTCH
LTE1.4MHz
GSMTCH
GSMTCH
GSMBCCH
GSMBCCH
GSMBCCH
Dynamic Spectrum Allocation and Interference Management
Active Antenna Systems High Mobility
SON Design Principles Multi-layer LTE/Wi-Fi TS
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
• In the remaining part of this presentation is on the Multi-layer LTE/Wi-Fi Traffic Steering use case, where the partners Ericsson, Nokia, iMinds and TNO have worked on.
• The following slides are an extented version from: http://www.fp7-semafour.eu/media/cms_page_media/20/6-
Wang-SEMAFOUR%20Workshop%202014-Traffic%20Steering.pdf
Thomas Kürner (TU Braunschweig) Slide 8
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Multi-Layer LTE / Wi-Fi Access Network Selection
Yu Wang+, Daniela Laselva*, István Z. Kovács*, Per-Henrik
Michaelsen*, Relja Djapicφ, Pieter Willemen^, Kathleen Spaeyx, Bart Sasx, Dries Naudts^, Andreas Bergström+
+Ericsson Research, Stockholm, Sweden; *Nokia, Aalborg, Denmark; φTNO, Delft, Netherlands; ^iMinds/Ghent University, Ghent, Belgium, x iMinds/University of Antwerp, Antwerp, Belgium
Thomas Kürner (TU Braunschweig) Slide 9
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Objective (1/2)
• Objective: – Access network selection between multi-layer LTE and Wi-Fi
in dense urban deployments to improve user experience and network efficiency
• Implemented SON functions: – Threshold based SON functions
• SON for LTE load control • SON for Inter-RAT LTE/Wi-Fi load control
– Throughput based SON functions • QoS-oriented access network selection, e.g. based on a per user
throughput metric
Thomas Kürner (TU Braunschweig) Slide 10
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Objective (2/2)
Thomas Kürner (TU Braunschweig) Slide 11
Average Load
Target
SON for LTE load control
Average Load
Target
SON for Inter-RAT LTE/Wi-Fi load control
Average Load
Target
Predicted LTE throughput
Throughput based SON function
Predicted Wi-Fi throughput
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Design – Monitoring KPIs
• LTE load – Average raw percentage of physical resource blocks (PRBs) utilization – Average fraction of required PRBs in a cell to serve connected UEs with a
certain minimum bit rate
• Wi-Fi load – Average percentage of channel busy time of a Wi-Fi AP – The channel is considered as busy if there is at least one active connection
associated to the AP and the AP or a UE is transmitting
How fast the SON functions can change configuration parameters which determine access network selection of UEs Observation & Adjustment period (0.5 – 2 seconds) Control parameter step size (0.5 – 5 dB)
Thomas Kürner (TU Braunschweig) Slide 12
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Evaluation - Scenario • Realistic dense urban environments • Dense WiFi deployment • Outdoor & Indoor
Thomas Kürner (TU Braunschweig) Slide 13
LTE Macro LTE Micro & WiFi User location Deployment: LTE macro layer: [email protected], 46dBm LTE micro layer: [email protected] 33dBm 6dB range extension WiFi 802.11n: [email protected]/5GHz, 20dBm
Traffic hot zone area
Outdoor Hot Zone
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Threshold based SON Functions
Thomas Kürner (TU Braunschweig) Slide 14
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Objective: LTE Load Control Objective: Inter-RAT Load Control
No
No
LTE load and WiFi load monitoring
LTE load is high
WiFi load is high?
Steer traffic to WiFi
LTE load is low
Steer traffic to LTE
Yes
Yes
No
No
No
No
LTE load and WiFi load monitoring
LTE load is high
WiFi load is high
Steer traffic to WiFi
LTE load is low
WiFi load is low
Steer traffic to LTE
Steer traffic to LTE
Steer traffic to WiFi
Yes
Yes
Yes
Yes
SON Function Design for the Access Network Selection
* When both LTE and WiFi load is high or low, no action will be taken
LTE load and WiFi load monitoring
Steer traffic to WiFi
Steer traffic to LTE
Steer traffic to LTE
Steer traffic to WiFi
Yes
Yes
Yes
Yes
No
No
No
No
SON for LTE load control
SON for Inter-RAT load control
WiFi load is high
LTE load is low
WiFi load is low
LTE load is high
?
?
Monitoring KPIs
Control Parameters
Thomas Kürner (TU Braunschweig) Slide 15
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Design – Control Parameters
Thomas Kürner (TU Braunschweig) Slide 16
&WiFi AP LTE Micro
&WiFi AP LTE Macro LTE Micro
RSS Threshold RSS Minimum Macro/Micro boundary
Macro
Micro
WiFi
LTE Macro
RSRP Threshold High RSS Minimum
Macro
Micro
WiFi
WiFi RSS Threshold LTE RSRP High & RSRP Low Thresholds
To steer more traffic to WiFi – Decrease WiFi RSS Threshold – Decrease RSRP High Threshold
Increase RSRP Low Threshold
Macro/Micro boundary RSRP Threshold Low
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Evaluation - Overview
• Evaluation domain:
LTE Load Control Inter-RAT Load Control
RSS Threshold RSRP Thresholds
Indoor Outdoor
Static Users Mobile Users
Objective
Control Parameter
Environment
Mobility
Presented in this presentation: – LTE Load Control + RSRP Thresholds + Outdoor + Static Users – Inter-RAT Load Control + RSS Threshold + Outdoor + Static Users
Thomas Kürner (TU Braunschweig) Slide 17
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Evaluation – LTE Load Control
• RSRP Threshold Low (macros) and RSRP Threshold High (micros) are properly adjusted to meet the objective, i.e. keep LTE cell load within target range (60-80%)
1 2 3 4 5 6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
10
20
30
40
50
60
70
80
90
100
Cell l
oad
(%)
LTE Macro Co-located LTE Micro / Wi-Fi AP
-110
-100
-90
-80
-70
-60
-50
-40
RSRP
Thr
(dBm
)
Cell LoadRSRP Thr HighRSRP Thr Low
Outdoor Hot Zone scenario without UE mobility
High load in macro#3 due to limited offloading possibilities
Thomas Kürner (TU Braunschweig) Slide 18
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Evaluation – LTE Load Control cont’
• The SON configuration results in 50% offload to Wi-Fi while good Wi-Fi throughput is achieved, i.e. 5th-ile throughput of 2.6 Mbps
• Overall good performance is achieved, avg UE throughput=19.1 Mbps – vs. Baseline “Wi-Fi if coverage” with mean UE throughput = 10.5 Mbps
• Micro performance lower than macro due to presence of cell-edge UEs Outdoor Hot Zone scenario without UE mobility
Thomas Kürner (TU Braunschweig) Slide 19
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Evaluation – Inter-RAT Load Control
• RSS Thresholds are properly adjusted to meet the objective, i.e. balance the load between LTE and Wi-Fi
1 2 3 4 5 1 2 3 4 5 6 7 8 9 1011121314151617180
20
40
60
80
100
Cel
l loa
d
-80-75-70-65-60-55-50-45-40
RS
S th
resh
old
(dB
m)
1 2 3 4 5 1 2 3 4 5 6 7 8 9 1011121314151617180
20
40
60
80
100
Cel
l loa
d
-80-75-70-65-60-55-50-45-40
RS
S th
resh
old
(dB
m)
Fixed Threshold = -65 dBm
SON function
Wi-Fi Load
LTE Load
RSS Threshold
Thomas Kürner (TU Braunschweig)
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Throughput-based SON Functions
Thomas Kürner (TU Braunschweig) Slide 21
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Principle of Throughput based Traffic Steering
• Throughput based Traffic Steering between LTE and Wi-Fi is based on predicted/measured user throughput in LTE and Wi-Fi
• A user session is served by the RAT (LTE or Wi-Fi) which provides the highest throughput (plus a hysteresis to minimize IRAT handovers)
• The algorithm runs @Call Setup and During a Call (if enabled)
Thomas Kürner (TU Braunschweig) Slide 22
Wi-Fi AP link (beacon and/or traffic)
WLAN Controller signalling link
LTE eNB signalling link
Inter-eNB X2 signalling link
LTE eNB traffic link
eNB SC1
eNB MC1
eNB SC2
Wi-Fi AP2
WLAN Controller
Data
Beacon detection
Wi-Fi AP1
LTE-WLAN interface exchange
(e.g. BSS Load) LTE/Wi-Fi
Throughput estimate and TS decision at eNB
LTE/Wi-Fi Throughput
estimate and TS decision at eNB
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Results from Throughput basd SON functions
• Throughput based SON function outperforms the threshold based functions – Gains of 25%-240% for the 5th percentile user throughput depending on scenarios – More balanced user throughput among the network layers
Thomas Kürner (TU Braunschweig) Slide 23
0 20 40 60 80 100 1200
50
100
CD
F [%
]DL user throughput (Mbps)
AllMacro (21%)Micro (25%)Wi-Fi (54%)
0
0.5
1
1.5
2
2.5
3
5th-
ile s
essi
on th
roug
hput
(Mbp
s)
RSRP-based12 UEs/AP
RSRP-basedInf UEs/AP
Tput-based@CallSetup
Tput-based@CS&DuringCall
LTE MacroLTE MicroLTEWiFiOverall
Throughput based SON function example
0 20 40 60 80 100 1200
50
100C
DF
[%]
DL user throughput (Mbps)
AllMacro (23%)Micro (24%)Wi-Fi (54%)
Outdoor dense urban with UE mobility Outdoor dense urban without UE mobility
RSS-based
Tput-based
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
SON Function Implementation
• The proposed SON functions are intended to be implemented in a distributed manner
– Control parameters are updated every few seconds
• Implementation of the access network selection rules – Executed in terminals assisted by the network: Control parameter thresholds are sent to
terminals via broadcasted or dedicated signalling channels being standardized in 3GPP (RAN2 R12)
– Controlled by the network: Control parameters are monitored at a network node and the node controls the access network selection
• Information exchange between LTE and WiFi – Standardization of such an interface is being discussed in 3GPP (Release 13 work item
RP151114 “LTE-WLAN Radio Level Integration and Interworking Enhancement” covers the LTE-WLAN exchange (between the LTE eNB and the WT (WLAN Termination function)) and the UE reporting of WLAN measurements required for implementing the presented schemes
– Proprietary interfaces – Terminals as relays
Thomas Kürner (TU Braunschweig) Slide 24
doc.: IEEE 802.11-15/0847r0
Submission
July 2015
Summary
• SON functions for LTE and WI-Fi traffic steering have been designed, evaluated and demonstrated in realistic dense urban scenarios
• The effectiveness of the SON functions was proved with controlled load levels and improved user throughput
• We found the performance of the SON functions were most sensitive to the control parameter updating pace, i.e. the step size and period
• Throughput-based SON outperforms threshold-based SON in user throughput and improved manageability at the cost of further complexity (throughput prediction)
• Among other findings, the study unveiled the importance of information exchange between LTE and Wi-Fi Thomas Kürner (TU Braunschweig) Slide 25
