FP7 ICT-SOCRATES Algorithms for Self- optimisation Presented by Ove Linnell (Ericsson) 9 th June 2009 ICT Mobile Summit 2009 Pre-conference Workshop on ‘Self-Organisation for Beyond 3G Wireless Networks’
FP7 ICT-SOCRATES
Algorithms for Self-
optimisation
Presented by Ove Linnell (Ericsson)
9th June 2009
ICT Mobile Summit 2009 Pre-conference Workshop on ‘Self-Organisation for Beyond
3G Wireless Networks’
WWW.FP7-SOCRATES.EU
2/20
Ove Linnell, Ericsson Research
Outline
�Self-optimisation in SOCRATES
�Self-optimisation principles
�Selected use cases– Self-optimisation of Home eNodeB
– Packet scheduling parameter optimisation
– Admission & Congestion control
– Load Balancing
– Interference Coordination
– Handover
� Interacting Self-Optimisation algorithms
�Summary / Conclusion
WWW.FP7-SOCRATES.EU
3/20
Ove Linnell, Ericsson Research
Self-Optimisation in SOCRATES
General definition:Optimise the performance of the network through automatic or semi-automatic features,
to save costs, running the networkto enhance the performance of the network
Measurements(Gathering and processing)
Self-optimisation
Settingparameters
Self-healing
Self-configuration
continuous
loop
triggered by incidental
events
Measurements(Gathering and processing)
Self-optimisation
Settingparameters
Self-healing
Self-configuration
continuous
loop
triggered by incidental
events
SOCRATES approach:
Develop new concepts, methods and algorithms for efficient self-optimisation of wireless networks that adapts to gradual changes in the radio network,
� with focus on LTE
� real-time optimisation on a time scale ranging from monthly or weekly to hours or minutes
� Optimise on a wider range of radio parameters(local optimisation � system optimisation)
� Quick response to network changes to resulting in optimisation that best suits current network situation
� Strong gains in coverage, capacity and quality of service, e.g. no need for drive tests.
Self-optimisation use cases in SOCRATES:
� Home eNodeB
� Packet scheduling
� Admission & congestion control
� Load Balancing
� Interference Coordination
� Handover
WWW.FP7-SOCRATES.EU
4/20
Ove Linnell, Ericsson Research
Network
Self-Optimization
Self-HealingSelf-Configuration
Observation
Action
Anomaly detection
Install
New
HW
Eternal cycle
Policy
Self-optimisation principles
WWW.FP7-SOCRATES.EU
5/20
Ove Linnell, Ericsson Research
� Aspects to consider
– How local is the problem? Single cell, regional or network-wide?
– How local is the impact of control actions?
– How isolated are the involved control parameters w.r.t. other mechanisms?
– How quick is a response needed? Within (milli)seconds, minutes?
– How much computational power is available?
– What degree of knowledge regarding the network is assumed by the algorithm?
– What is the impact on network during learning phase?
� A way forward when selecting Self-optimisation principles,
– First get it quickly roughly right off-line using
– An optimisation tool
– Measurement-tuned propagation maps are used
– Expert-based rule of thumb
– Then fine-tune control parameters
– On-line optimised real-time controller that is allowed only gradual
parameter adjustments within a limited range
Self-optimisation principles
WWW.FP7-SOCRATES.EU
6/20
Ove Linnell, Ericsson Research
Self-optimisation of Home eNodeB
� Up to 70% of the traffic load is indoor
� Support open and closed subscriber groups
� Minimal human intervention
� Selected sub-use cases
– HO optimisation
– HO between macro and Home eNodeB should work smoothly and
seamless
– Probably the user wants to move to Home eNodeB as quick as
possible and stay there as long as possible
– Interference and Coverage Optimisation
– Minimize interference to macro in order not to deteriorate macro
performance
– Avoid coverage holes
WWW.FP7-SOCRATES.EU
7/20
Ove Linnell, Ericsson Research
Packet scheduling parameter optimisation
� Appropriately tune the parameters of the packet scheduling algorithm in response to changes,
– in e.g., traffic load, traffic mix, traffic characteristics, interference
characteristics, etc.
– such that channel resources are optimally used
– such that a balanced QoS is offered to all users
� Reference algorithm selected
� The focus is on the sensitivity of the optimum parameter settings of the scheduler with scenarios like :
– data only
– spatially inhomogeneous traffic (hot spots)
– mobility (sudden changes in a user’s channel condition, e.g. due to
shadowing)
– mixed traffic: video/data
WWW.FP7-SOCRATES.EU
8/20
Ove Linnell, Ericsson Research
Admission & Congestion control
� Self-optimise the admission & congestion control parameters in response to changes in cell capacity, traffic mix etc. so that QoS of admitted calls is meet and that call rejection is minimized
� Admission control
– Admits or rejects calls requests
– Proactive mechanism to prevent undesired/uncontrolled QoS
degradation
� Congestion control
– Reactive mechanism once QoS degradation has occurred
� Reference algorithm selected
– Focus on admission control for DL
– Priority will be made for HO calls rather than for new calls
– To ensure the QoS requirements priority will be given to real-time calls
instead of non- real time calls
WWW.FP7-SOCRATES.EU
9/20
Ove Linnell, Ericsson Research
Other Self-optimisation use cases
�Load Balancing
– HO Offset modifications
– Cell coverage adjustments
� Interference Coordination
– Impact of different environments, service type, user mobility, traffic load distribution and elevated UE:s
�Handover
– Optimal setting of HO parameters
– Avoid ping-pong effects, e.g. in the stationary scenario
– Minimize HO drops
WWW.FP7-SOCRATES.EU
10/20
Ove Linnell, Ericsson Research
Interacting Self-Optimisation algorithms 1(2)
� Each Self-Optimisation (SON) algorithm have one or several goals to meet
� A SON algorithm alter a set of control parameters to reach a predefined goal, e.g. KPI
� Coordination needed because,
– Algorithms have conflicting goals
– Could alter same control parameters
Goal 1 Goal 3Goal 2
SONAlgorithm 1
SONAlgorithm 2
SONAlgorithm 3
SONAlgorithm 4
…
…
Ctrl. param 1 Ctrl. param 2 Ctrl. param m…
Coverage, capacity, quality
optimizationBalance load
Optimize HO performance
Antenna
parameters
Transmit power HO parameters
WWW.FP7-SOCRATES.EU
11/20
Ove Linnell, Ericsson Research
Interacting Self-Optimisation algorithms 2(2)
Goal 1
Coordinator
Goal 3Goal 2
SONAlgorithm 1
SONAlgoritm 2
SONAlgorithm 3
SONAlgorithm 4
…
…
Ctrl. param 1 Ctrl. param 2 Ctrl. param m…Operatorpolicy
� The SON algorithms may specify
– desired changes to control parameters
– desired range of the control parameters
� The coordinator may decline desired change or partially accept change by taking into account
– desired changes of all SON algorithm
– impact on goals and/or operatory policies
� Using the interface, the coordinator informs the SON algorithm regarding, e.g.,
– if desired change cannot be accomodated
– changes to other parameters that may impact the goal of the SON function
WWW.FP7-SOCRATES.EU
12/20
Ove Linnell, Ericsson Research
Summary / Conlusion
� Introducing Self-optimising features to save cost, but also able to cope with the increasing complexity
� Some Self-optimisation principles presented
� Six uses cases selected for further elaboration– Self-optimisation of Home eNodeB
– Packet scheduling parameter optimisation
– Admission & Congestion control
– Load Balancing
– Interference Coordination
– Handover
� Challenging to cope with interacting (conflicting?) Self-optimisation algorithms
– Addressed within next phase of SOCRATES