Philipp Hasselbach Capacity Optimization for Self- organizing Networks: Analysis and Algorithms Philipp Hasselbach
2 Inhomogeneous capacity demand Rush hour traffic Concerts, sports tournaments Change in user behaviour Motivation Capacity Optimization As much capacity as required At all times and all places Achieved by allocation of cell bandwidth and transmit power to the cells
3 Philipp Hasselbach Capacity in Cellular Networks Downlink considered Link capacity influencing factors User position Attenuation Shadowing Inter-cell interference Cell capacity influencing factors User distribution Service type Scheduling Transmit power Cell bandwidth Inter-cell inter- ference power SINR of user k Noise power
4 Philipp Hasselbach Self-organizing Networks (SONs) Drivers High complexity of mobile radio technology Operation of several networks of different technologies Need to reduce CAPEX and OPEX Autonomous operation In configuration, optimization, healing Circumventing classical planning and optimization processes SONS: Shift of paradigm Source: FP7 SOCRATES
5 Philipp Hasselbach Automatic Capacity Optimization for SONs Real-time capabilities Treatment of large networks Accurate results Reliable operation Depends on user distribution environment Inter-cell interference (ICI) Interdependencies among cells and users Capacity optimizationSON requirements Source: FP7 SOCRATES High complexity, excessive signaling
6 Philipp Hasselbach Outline Cell-centric Network Model Requirements and Derivation PBR- and PBN-Characteristic Automatic Capacity Optimization for SONs Self-Organizing Approach Network State evaluation Network Capacity Optimization Simulation and Analysis Functional Analysis Real-World Analysis Summary
7 Philipp Hasselbach Outline Cell-centric Network Model Requirements and Derivation PBR- and PBN-Characteristic Automatic Capacity Optimization for SONs Self-Organizing Approach Network State evaluation Network Capacity Optimization Simulation and Analysis Functional Analysis Real-World Analysis Summary
8 Philipp Hasselbach Cell-centric Network Model: Requirements Application for allocation of resources cell bandwidth and transmit powers to the cells Modeling of the relation between cell bandwidth, transmit power and cell performance Low complexity Consideration of User QoS requirements User distribution Environment Inter-cell interference Interdependencies among cells
9 Philipp Hasselbach Cell-centric Network Model User bit rate Cell throughput PBR-Characteristic SINR measurements User distribution, environment model
10 Philipp Hasselbach Cell-centric Network Model User bit rate Number of users User QoS requirements Cell throughput PBR-Characteristic User bit rate pdf empiric theoretic SINR measurements User distribution, environment model
11 Philipp Hasselbach Cell-centric Network Model Cell throughput in Mbit/s User bit rate Number of users User QoS requirements Cell throughput PBR-Characteristic Outage probability p Cell bandwidth B Transmit power P User bit rate pdf empiric theoretic Cell throughput cdf empiric theoretic SINR measurements User distribution, environment model p
12 Philipp Hasselbach PBR- and PBN-Characteristic PBR-Characteristic Relates transmit power, cell bandwidth, cell throughput of cell i PBN-Characteristic Relates transmit power, cell bandwidth, number of users of cell i Cell throughput in Mbit/s Number of users
13 Philipp Hasselbach PBR- and PBN-Characteristic PBR-Characteristic Relates transmit power, cell bandwidth, cell throughput of cell i PBN-Characteristic Relates transmit power, cell bandwidth, number of users of cell i Cell throughput in Mbit/s Number of users Power ratio: relates transmit power to average inter-cell interference power
14 Philipp Hasselbach PBR- and PBN-Characteristic PBR-Characteristic Relates transmit power, cell bandwidth, cell throughput of cell i PBN-Characteristic Relates transmit power, cell bandwidth, number of users of cell i Cell throughput in Mbit/s Number of users Available for different schedulers
15 Philipp Hasselbach Outline Cell-centric Network Model Requirements and Derivation PBR- and PBN-Characteristic Automatic Capacity Optimization for SONs Self-Organizing Approach Network State evaluation Network Capacity Optimization Simulation and Analysis Functional Analysis Real-World Analysis Summary
16 Philipp Hasselbach Self-organizing Approach Self-organizing control loop: Network state optimization Application of PBR-/PBN-Characteristic Determination of possible performance Comparison with required performance Decision to take action Network capacity optimization Definition of optimization problems Application of PBR-/PBN-Characteristic in objective function and constraints Solution of optimization problems to obtain resource allocation to cells Constant cell sizes Cell throughput in Mbit/s Collection of measure- ments Network state evaluation Network capacity optimisation Cellular radio network
17 Philipp Hasselbach Network State Evaluation Current network state Number of users in cell i: Cell bandwidth: Power ratio: Number of users that can be supported by the cell (obtained from PBN-Characteristic): : no action : network optimization Number of users
18 Philipp Hasselbach Network Capacity Optimization Network through- put R Total number of users N Transmit power P Cell band- width B Joint P,B Optimization problemsOptimization approaches
19 Philipp Hasselbach Network Capacity Optimization Network through- put R Total number of users N Transmit power P Cell band- width B Joint P,B Optimization problemsOptimization approaches Central and distributed solving algorithms for analysis and implementation
20 Philipp Hasselbach Outline Cell-centric Network Model Requirements and Derivation PBR- and PBN-Characteristic Automatic Capacity Optimization for SONs Self-Organizing Approach Network State evaluation Network Capacity Optimization Simulation and Analysis Functional Analysis Real-World Analysis Summary
21 Philipp Hasselbach Simulation Approach for Functional Analysis Inhomogeneous capacity demand: hotspot scenarios users in hotspot cell, users in non-hotspot cell Hotspot factor Wrap-around technique to avoid border effects Evaluation of capacity optimization approaches w.r.t. hotspot distribution Evaluation for different hotspot strengths w/o coordination of bandwidth allocations of neighbored cells Mitigation of inter-cell interference LTE-typical simulation parameters Single hotspot scenario Cluster hotspot scenario Multi hotspot scenario
22 Philipp Hasselbach Simulation Parameters for Functional Analysis Cell radius R250 m Number of cells39 User distributionuniform Propagation model3GPP SCM Urban Macro Shadow fading variance8 dB Max. transmit power33 dBm Total system bandwidth10 MHz SchedulingPF, FT Data rate per user100 kbit/s
23 Philipp Hasselbach Network Throughput Optimization, Single Hotspot Scenario PF scheduling FT scheduling
24 Philipp Hasselbach Network Throughput Optimization, Coordinated Bandwidth Allocations Cluster HS Scenario Multi HS Scenario
25 Philipp Hasselbach Functional Analysis: Summary Adaptation of the network to inhomogeneous capacity demands achieved For strong inhomogeneous capacity demand coordination of bandwidth allocations required For FT scheduling coordination of bandwidth allocations required Transmit power allocation favorable with clustered hotspot cells Cell bandwidth allocation and joint allocation favorable with distributed hotspot cells
26 Philipp Hasselbach Simulation Approach for Real-World Analysis Scenario based on real network Network footprint from existing network Downtown area, 50 km², 46 sites, 126 sectors Pilot power receive strength prediction for each sector Determination of cell borders Inhomogeneous capacity demand According to user distribution estimation Based on DL throughput measurements 229 snapshots over 5 days Performance analysis Consideration of snapshots Evaluation of performance in strongest hotspots
27 Philipp Hasselbach Real-World-Analysis: Hotspot Strength and Strongest Hotspots Strongest hotspot Maximum hotspot strength
28 Philipp Hasselbach Real-World-Analysis: Hotspot Strength and Strongest Hotspots Strongest hotspot Maximum hotspot strength
29 Philipp Hasselbach Real-World-Analysis: Hotspot Strength and Strongest Hotspots Strongest hotspot Network throughput, FT scheduling
30 Philipp Hasselbach Outline Cell-centric Network Model Requirements and Derivation PBR- and PBN-Characteristic Automatic Capacity Optimization for SONs Self-Organizing Approach Network State evaluation Network Capacity Optimization Simulation and Analysis Functional Analysis Real-World Analysis Summary
31 Philipp Hasselbach Summary Cell-centric network modeling proposed PBR- and PBN-Characteristic Provides accurate modeling for automatic capacity optimization for SONs Avoids high complexity and high signaling effort Self-Organizing Approach proposed Application of cell-centric network model Central and distributed implementations for analysis and practical implementation Simulative verification In artificial scenarios and real-world scenario Adaptation of the network to inhomogeneous capacity demands shown
32 Philipp Hasselbach Backup
33 Philipp Hasselbach Power-Bandwidth Characteristics User distribution PDF of the bandwidth required by user k Bandwidth required by user k Bandwidth required by the whole cell PDF of the bandwidth required by the cell K independent users Central Limit Theorem
34 Philipp Hasselbach Cell Outage Probability CDF of the bandwidth required by the cell Probability that sufficient bandwidth is allocated Cell outage probability Probability that allocated bandwidth is not sufficient Bandwidth required by the cell
35 Philipp Hasselbach
36 Philipp Hasselbach Fluctuating capacity demand Rush hour traffic Concerts, sports tournaments Change in user behaviour Change in environment Motivation Capacity Optimization As much capacity as required At all times and all places
37 Philipp Hasselbach Automatic Capacity Optimization for SONs Real-time capabilities Accurate results Reliable operation Complex modeling Large number of users and BSs Effects of the user distribution Effects of the environment Interdependencies among cells and users Source: FP7 SOCRATES Capacity optimizationSONs
38 Philipp Hasselbach Automatic Capacity Optimization for SONs Real-time capabilities Accurate results Reliable operation Complex modeling Effects of the user distribution Effects of the environment Inter-cell interference (ICI) Interdependencies among cells and users Capacity optimizationSONs Source: FP7 SOCRATES
39 Philipp Hasselbach
40 Philipp Hasselbach Cell-centric Network Model User distribution, environment model SINR measurements Outage probability Cell bandwidth Transmit power User QoS requirements Cell throughput in Mbit/s
41 Philipp Hasselbach Cell-centric Network Model Outage probability Cell bandwidth B Transmit power P User distribution, environment model SINR measurements User bit rate pdf empiric theoretic Number of users User QoS requirements Cell Performance for (B,P)
42 Philipp Hasselbach PBR-Characteristic Reduced complexity due to focus on cells User QoS requirements considered Relation between cell bandwidth, transmit power and cell performance Cell throughput in Mbit/s Cell Performance for (B,P) For different Cell bandwidth B Transmit power P
43 Philipp Hasselbach
44 Philipp Hasselbach Model the interdependence of transmit power and cell bandwidth Contain information on user distribution, environment, inter-cell interference Analytic derivation available Measurement based derivation available, determined from standard system measurements (attenuation, SINR) Cell-centric Network Model Modeling equations Random Variable transformation Measurement data transformation SINR measurements User distribution Environment model Theoretic Approach Practical Approach
45 Philipp Hasselbach Cell-centric Network Model User distribution, environment model SINR measurements Number of users Outage definition Cell bandwidth Transmit power
46 Philipp Hasselbach
47 Philipp Hasselbach Automatic Capacity Optimization Approaches I take SC 1. OK, I take SC 2 Can I take SC 1? Uncoordinated/scheduling based (State of the art): Coordinated (new): + : easy implementation - : Collisions, QoS? + : Collisions can be avoided QoS - : Complexity? Implementation? I take SC 1. Local Scheduling Inter-BS communication SC1 SC2
48 Philipp Hasselbach Two Alternative SO Approaches I take SC 1. OK, I take SC 2 Can I take SC 1? Uncoordinated:Coordinated: I take SC 1. Local Scheduling Inter-BS communication SC1 Power-Bandwidth Characteristic for performance analysis Power-Bandwidth Characteristic for approach realization and per- formance analysis
49 Philipp Hasselbach General System Concept Hierarchical approach Resource allocation to users, no inter-cell scheduling Sched. cell 1 Sched. cell 2 Sched. cell N Resource allocation to cells Source: 3GPP Network state evaluation Network parameter optimisation Network parameter adjustment Self-organising functionality/ Self-organising control loop