Admission Control Optimisation and its Influence on Handover Optimisation

Transcription

1 FP7 ICT-SOCRATES Admission Control Optimisation and its Influence on Handover Optimisation Presented by Bart Sas IBBT FP7 SOCRATES final workshop Karlsruhe, Germany 22 February 2011 Contributors: Kathleen Spaey (IBBT), Irina Balan (IBBT), Kristina Zetterberg (EAB), Remco Litjens (TNO ICT)

2 Outline Introduction Admission control Handover Goal Scenarios Expected conflicts Overview Simulation environment Approach Simulation results Conclusions More details can be found in D5.9, section 9.3 2

3 Introduction 3

4 Admission control (AC) Determines whether a call is admitted to a cell or not In order to guarantee QoS Distinguishes between fresh and handover (HO) calls HO calls are given priority over fresh calls Dropped calls are a greater nuisance for users than rejected fresh calls Determined by the Th HO parameter C(t) Reserved for HO calls 100 % Th HO reduce rejection fresh reduce rejection HO 0 % 4

5 Admission control parameter optimisation (AC SON) AC parameter optimisation algorithm tunes the Th HO parameter In order to adapt the AC algorithm to changes in the environment Collect KPIs: Rejection ratio of the fresh calls (RRFC) Rejection ratio of the HO calls (RRHOC) Low throughput ratio (LTR) Traffic loss ratio (TLR) Optimise: If RRHOC, LTR or TLR > Threshold: lower Th HO Otherwise, if RRFC > Threshold: raise Th HO 5

6 Handover (HO) Determines when calls that move away from a cell are handed over to another cell In order to maintain connectivity The moment when a call is handed over is determined by 2 parameters Hysteresis: the minimum difference between the signal strength of the TenB and the SeNB Time-to-trigger: the amount of time before the handover is triggered 6

7 Handover parameter optimisation (HO SON) HO optimisation algorithm tunes TTT and hysteresis In order to adapt the HO algorithm to changes in the environment Collect KPIs: Call drop ratio (CDR) Handover failure ratio (HOFR) Ping pong handover ratio (PPHOR) Optimise: If current HPI > previous HPI: change optimisation direction Adapt Hys/TTT in optimisation direction HPI = 2 * CDR + 1 * HOFR * PPHOR 7

8 Goal of the AC and HO SON integration Admission control (AC) and handover (HO) optimisation algorithms were developed separately In reality both algorithms will have to operate in parallel In this case both algorithms might influence each other Goal of the AC and HO SON integration use case Deploy both AC and HO optimisation algorithms together Study interaction between both optimisation algorithms Resolve possible conflicts AC SON Adapts parameters: Th HO HO SON Adapts parameters: Hys, TTT Simultaneous operation: Interaction? Conflicts? Need for integration? 8

9 Scenarios 9

10 Expected conflicts Conflicts between the AC and HO SON algorithms might occur when the AC algorithm rejects a lot of HO calls Because of overload Results in high number of call drops Because calls do not find a HO target Overload causes many HO calls to be dropped HO SON might react Because it thinks the HO parameters are wrong Reaction is not desired Instead, AC SON should react By resolving overload HO SON will react AC SON should react 10

11 Scenarios # Change Velocity Load Before After Before After 1 Gradual 3 km/h 50 km/h 2% RRFC 2 Abrupt 3 km/h 50 km/h 2% RRFC 3 Gradual 3 km/h 2% RRFC 20% RRFC 4 Abrupt 3 km/h 2% RRFC 20% RRFC 5 Gradual 3 km/h 50 km/h 2% RRFC 20% RRFC 6 Abrupt 3 km/h 50 km/h 2% RRFC 20% RRFC Evaluate how one of the algorithms affects the targets of the other Evaluate how the two algorithms work together The impact of two different parameters is considered UE velocity: influence on HO SON Traffic load: influence on AC SON A A B B 11

12 Simulator overview Dynamic system-level simulator Downlink direction is simulated Network layout: 25 cells in a 5x5 grid 500 m site-to-site distance Propagation model: Okumura-Hata for large urban areas pathloss model Both auto- and cross correlated lognormal shadow fading Call and traffic generation: 2500 users generate calls according to a Poisson process Real time and non-real time traffic Mobility model: Random walk 12

13 Approach Scenarios are simulated in four different ways: Without any SON algorithm enabled With only the AC SON algorithm enabled With only the HO SON algorithm enabled With both the AC and HO SON algorithms enabled AC SON HO SON Motivation: Observe differences between cases Deduce influences of the algorithms on each other 13

14 Simulation results 14

15 Gradual speed and load change rejection ratio of handover calls Rejected Handover Calls No SON AC SON HO SON AC & HO SON Before After RRHOC is lower if AC SON is enabled

16 Gradual speed and load change rejection ratio of fresh calls Rejected Fresh Calls No SON AC SON HO SON AC & HO SON Before After As a consequence the RRFC is higher

17 Gradual speed and load change call drop ratio Call Drops No SON AC SON HO SON AC & HO SON Before After Cases in which AC SON is enabled outperform their counterparts without AC SON If the HO SON algorithm is enabled the CDR is lower

18 Gradual speed and load change ping pong handover ratio Ping Pong Handovers Before After No SON AC SON HO SON AC & HO SON As a trade-off the PPHOR is higher The lower CDR in case the AC SON is enabled does not cause a higher PPHOR

19 Conclusions 19

20 Conclusions Scenarios with various changes (speed, load) were defined and run The HO SON lowers the CDR at the expense of the PPHOR The AC SON lowers the RRHOC at the expense of the RRFC The reduced number of rejected handover calls when AC SON is enabled has a positive influence on the CDR Calls will be accepted and handed over more rapidly HO calls will penetrate less far into target cell Signal quality degrades less Fewer calls will be dropped No additional increase of PPHOR Expected interaction does not occur Because AC SON resolves the problem fast enough Interaction might be possible with other implementations 20

21 Questions? 21

22 Gradual speed change call drops Call Drops No SON AC SON HO SON AC & HO SON Scenarios in which HO SON is enabled perform better on CDR

23 Gradual speed change ping pong handover ratio Ping Pong Handovers As a trade-off the PPHOR is higher No SON AC SON HO SON AC & HO SON

24 Gradual speed change rejected fresh calls Rejected Fresh Calls No SON AC SON HO SON AC & HO SON The RRFC is lower because the load is lower due to the higher CDR

25 Speed change summary The HO SON lowers the CDR As a trade off the PPHOR is higher The AC SON does not influence the results The same conclusions are valid for abrupt changes 25

26 Gradual load change rejection ratio of the handover calls Rejected Handover Calls No SON AC SON HO SON AC & HO SON RRHOC is lower if AC SON is enabled

27 Gradual load change rejection ratio of the fresh calls Rejected Fresh Calls No SON AC SON HO SON AC & HO SON As a consequence the RRFC is higher

28 Gradual load change call drop ratio The CDR is not high Call Drops No SON AC SON HO SON AC & HO SON The CDR is not influenced by the load change

29 Gradual load change ping pong handover ratio The PPHOR is not influenced by the load change Ping Pong Handovers No SON AC SON HO SON AC & HO SON

30 Load change summary The AC SON lowers the RRHOC As a trade off the RRFC is higher The HO SON does not influence the results The same conclusions are valid for abrupt changes 30