1 Amit Mukhopadhyay, Ph. D. Director, Network Modeling & Optimization - Wireless Bell Labs Self-Organizing Networks in LTE and Quantification of Economic Benefits
1
Amit Mukhopadhyay, Ph. D.
Director, Network Modeling & Optimization - Wireless
Bell Labs
Self-Organizing Networks in LTE and Quantification
of Economic Benefits
2
Agenda
• Self Organizing Networks (SON) fundamentals
• Economic model
3
4
Self Organizing Network (SON) is a 3GPP-standardized initiative which drives the
building of intelligence and automation into the network itself to help operators solve
those challenges
SON provides Operators with a path to maximize their network performance with
lower effort (and thus lower cost).
3GPP SON helping Operators solve their challenges
Increase network quality
OPEX reduction
Fast adaptation to network conditions
• Provide Higher End User Quality of Experience
• Ensure service continuity
• Embed system and product expertise in the network
• Avoid time-consuming and repetitive tasks
• Avoid drive tests
• Process simplification
• Avoid error-prone and slow manual operations
• world of data is dynamic so networks must adapt real time
5
EPS Architecture: eUTRAN and EPC Network Topology
Standards described the EPC elements as logical functions that can be integrated or distributed
- Physical topology & implementation varies based
on numerous considerations
Standards specified that eNBs are inter-connected via the X2 reference point to improve handover performance
- X2 connectivity is logical, not necessarily
physical
- Many options of creating logical connectivity, with
different cost implications
eNB typically connects to multiple MME’s and S-GW to leverage load sharing / balancing capability and increases reliability
MME
eNB eNB
eNB
MME
SGW
X2
X2X2
SGW
PCRF
PDN GW
PDN GW
S1-MME S1-U
S1-US1-MME
S1-U
S1-MME
S11
S7c
S5/S8
S7
eUTRAN
EPC
6
SON Overview
� Self Configuration :
• Automatic installation procedures to get the necessary basic configuration for system operation.
• Works when the eNB is powered up and has backbone connectivity but the RF transmitter is not yet switched on.
• Functions: Basic Set up + Initial Radio Configuration
�Self Optimization
• UE & eNB measurements and performance measurements are used to auto-tune the network.
• Works when the pre-operational state is complete and the RF interface is additionally switched on.
• Function: Optimization/Adaptation
Fig 22-1-1: 3GPP TS 36.300
7
Example Prelaunch and Post-launch Optimizations
7
Design Configure
Drive Test
Analyze DT Results Resolve/Reconfigure
Pre-Launch
PM Collect Data (Remotely)
PM Analyze
Top 10
Analyze DT Results
Resolve/Reconfigure
Post-Launch – Continuous operations
RFBackhaulOther
Don’t know/DT requiredField Visits Remote ly Resolvable
Drive Test
Repeat until desired configurations achieved
8
Key SON features
• Self Configuration
- Network connection, database update, software download,…
- Physical Cell ID configuration
- Neighbor Recognition
• Self Optimization
- Load balancing
- Hand-off optimizatyion
- Interference co-ordination
- Capacity and coverage optimization
- Energy optimization
• Self Healing
- Outage detection and compensation
- Multi-homing
9
Physical Cell ID configuration
• Cell PCI needs to be:
- Collision free
- Confusion free
• Only 504 unique PCIs available, which need to be recycled among cells
• Determined during planning via planning tool and then configured into eNodeB via EMS or manually/config files
New cell
PCI
algo
Automate configuration efforts
10
B
A
Automatic Neighbor Recognition (ANR)
As a part of the normal call procedure, the eNB instructs each UE to perform measurements on neighbor cells.
1.The UE sends a measurement report regarding cell B. This report contains Cell B’s PCI.
2.The eNB instructs the UE, using the newly discovered PCI as parameter, to read the ECGI
3.When the UE has found out the new cell’s ECGI, the UE reports the detected ECGI to the serving cell eNB.
4.The eNB decides to add this neighbor relation
Reduce pre-launch and post-launch configuration efforts
11
LTE Load Balancing
� Objective: Distribute UEs camping on or connected to a cell to balance the traffic load.
� Delay or advance the handing over of UEs between cells.
� eNB monitors load in the cell and exchanges related information over X2 or S1 with neighbouring node(s)
Load Balancing
Distribute load across neighbouring cells
12
LTE SON Vision:
Inter-RAT Load Balancing for Layered Heterogeneous Networks
1. 3G-only UE: stays on 3G as it moves through network
2. 3G+LTE UE: moves between 3G macro, LTE macro, & LTE macro/pico
3. 3G+LTE UE: able to access 3G & LTE macros, plus LTE HeNB (femto)
LTE Macro
Partial LTE deployment
(midterm transition stage)
3G Macro
Ubiquitous 3G deployment
GeographicLayout
LTE Pico + Femto Femto
Pico Femto
12
3
13
Robustness Optimization
13
• Objective: detecting and enabling correction of following problems:
- Connection failure due to intra-LTE mobility
- Unnecessary HO to another RAT
• HO failure categories
- Failures due to too late HO triggering
- Failures due to too early HO triggering
• Requirements:
- Detection of too late/early HO or wrong cell HO
- Reducing inefficient use of network resources due to unnecessary HOs
Identify failed/unnecessary HOs and suggests changes in HO parameters
14
Coverage and Capacity Optimization
• Objective: Users’ ability to establish and maintain connections with acceptable or default service quality
• Implies:
- Coverage is continuous
- Users are unaware of cell borders
• Coverage includes both idle and active modes for both UL and DL
• Coverage optimization has higher priority than capacity optimization
• Optimization of parameters through analysis of call trace data
• Also used for drive test minimization
- Requires UEs with GPS and data logging
- Privacy concerns “not applicable”
Trade-off between capacity and coverage
15
Inter Cell Interference Coordination (ICIC)
- Implemented to improve interference-limitations found in Cellular systems deployed with universal reuse
- Inter Cell Interference Coordination
- Uplink – Soft Fractional Frequency Reuse, Fractional Frequency Reuse
- Downlink - Power limitation on frequency blocks
- Improvements in later releases
- Antenna parameter optimisation e.g. tilt, azimuth
Improve user throughput and customer experience
16
Energy Consumption Optimisation
• Drivers towards a Greener network – Reduce Carbon Footprint
• Full eNode B switch-off is not possible: idle mode insteadResponsiveness � Component level (CPU, FPGA)
� Board level (modem)
� Cell level (Carrier)
� Site level (Node B)
Reduce Power consumption; additional benefit - Interference Reduction
17
Automatic Cell Outage Detection and Compensation
Two tasks:
�Detect Cell Outage
� In eNB with mis-operating cell
� In surrounding cells of neighbour eNBs
�Compensate for Cell Outage
� Fault report to OA&M
� From affected eNB
� From neighbour eNBs
� Reconfig of antenna parameters
� Azimuth and tilt adjustment via RET
� Reconfig of neighbour power
� Coordination between affected eNBs
Minimize system capacity loss following outage
18
Interactions between various SON use cases
Self configurationICIC
Hand Over
optimization
Load Balancing
optimization
Capacity
Optimization
Energy
Savings
RACH
optimization Cell Outage
detection/comp
Coverage
Optimization
Automatic
Neighbor Relation
Physical Cell ID
configuration
Terrain
OPEX impact
Coverage impact
Capacity impact
LEGEND
2G/3G/4G
Use case Interactions must be modeled and quantified to realize full SON potential
19
20
Lifecycle Cost Impacts of SON
Network Planning
Site Acquisition &
Preparation
Physical Installation
Commissioning
Integration
Pre Launch RF
Optimization
OA & M
Indicates SON impact
Pre-Launch Activities: Onetime CapEx
Post-Launch Activities: Recurring OpEx
Several Pre- and Post-launch Activities Leverage SON Efficiencies
21
CapEx Elements Impacted by SON
RAN
Core
Transport
Equipment
Deployment Services
Site Construction
Services not impacted
by SON features
Services impacted
by SON features
Study
focus
Network CapEx RAN CapEx Deployment Services CapEx
A Relatively Small Portion of Network CapEx is Impacted by SON
22
OpEx Elements Impacted by SON
CPGACPGA
CCPU
CCPU
CPGACPGA
CCPU
CCPU
Study
focus
A Relatively Small Portion of Network OpEx is Impacted by SON
23
Mapping of Key SON Features to Business Impacts
Note: Time to Market (TTM) and Churn Reduction impacts have been left out
24
Computation of Key Cost Elements
Note: Only CapEx and OpEx included in this analysis, not Marketing Costs
25
Key Modeling Assumptions
• Effort estimate parameters:
− Scheduled over a 5 year period implementing > 10,000 sites
− 2.5 million subscribers at the end of 5 years
− ~ 90% of current 2G/3G cell sites reused
• Unit labor cost estimate parameters:
− Typical wireless operator requires 900 and 1200 employees to manage RAN capacity and service assurance,
− Average annual loaded salaries of ~ $100,000
• Energy cost estimate parameters:
- Cell sites in business centers and dense urban areas are often lightly loaded
- 10% of sites in these areas could be powered down overnight
25
26
CapEx and OpEx Savings Components
Relative % reduction of CapEx Activities due to SON
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
Network Planning Commissioning Integration Pre-launch RF
Optimization
% CapEx Without SON
%CapEx With SON
Relative % reduction of OpEx Activities due to SON
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
Power OA&M
% OpEx Without SON
%OpEx With SON
27
Concluding Remarks
60.37%47.59%
39.63%
25.99%
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
Without SON With SON
Capex % Opex %
•A relatively small part of Network CapEx and OpEx benefit directly from SON; but cost areas where SON is impactful, the benefits can be significant
•SON benefits for an operator scales linearly with the size of the network; future mini/pico cells may bring benefits of SON in a more significant way
•Impacts of SON are also reflected in faster deployment and better QoE which translate to Time To Market as well as churn reduction benefits
•Future SON and xSON features may open the doors for unforeseen benefits.
28
High Traffic Density Area
Serving Cell/Sector
Normal Traffic Map
γ α
β
γ α
β
γ α
β
γ α
β
γ α
β
γ α
β
γ α
β
BS I BS III
BS II
γ α
β
γ
β
γ α
β
γ α
β
BS I BS III
BS II
γ α
β
γ α
β
α
γ α
β
Serving Cell/Sector
Has Outage
BS II shared load
A Final Thought: Outage Detection and Self Healing
γ α
β
γ
β
γ α
β
γ α
β
BS I BS III
BS II
γ α
β
γ α
β
α
γ α
β
Dynamically minimize service interruption caused by cell outage based on real timetraffic loading. Adjust antenna configuration in nearby cells.
High Traffic Density Area
Load is shared by
The Neighboring cells BS I, II & III
⇒Minimize impact of cell outage
29
Acknowledgments
• Craig Connick, Kamakshi Sridhar, Susan Sanders, Louise Gabriel, Ashoke Sharma and many other colleagues at Alcatel-Lucent