1. Group 896 GPRS Radio Resource Management Allocation of Radio Resources to packet-switched traffic in GPRS-like networks By group 896 Supervisors:Students: Hans Peter SchwefelXin Zhou Patrick Eggers Jimena Llorente Martinez Devendra Prasad François Delawarde Gwénaël Coget Haibo Wang

2. Group 896 GPRS Radio Resource Management Project Description: Analyze existing GPRS radio resource management strategies. Define scenarios and input parameters, for which an algorithm for the assignment of resources to packet-switch traffic will be developed. Evaluate/validate the algorithm in simplified simulation models.

3. Group 896 GPRS Radio Resource Management GPRS Introduction: Reuse the existing GSM infrastructure Introduce packet-switched routing functionality Better data transfer rates (multislot capabilities) Low cost and connectivity-oriented Migration Path to 3G Networks

4. Group 896 GPRS Radio Resource Management Comparison between GSM/GPRS

5. Group 896 GPRS Radio Resource Management GPRS architecture

6. Group 896 GPRS Radio Resource Management GPRS protocols stack

7. Group 896 GPRS Radio Resource Management GPRS Protocols:  GSM-RF layer:  Reuse of GSM multiple access methods (TDMA/FDMA)  MAC layer:  Multiplexing  Scheduling  Contention resolution (UL)  RLC layer:  Segmentation/Re-assembly  ARQ

8. Group 896 GPRS Radio Resource Management Radio Resource Management (RRM) Introduction -Functions GSM/GPRS establishment GSM/GPRS call maintaining GSM/GPRS release -Dedicated/Shared channels Best Effort Strategy Our Proposed RRM Strategy Resource Allocation Reassignment Scheduling

9. Group 896 GPRS Radio Resource Management Best Effort : Access Control

10. Group 896 GPRS Radio Resource Management EXAMPLE: Resource Assignment Best Effort

11. Group 896 GPRS Radio Resource Management Best Effort : Round Robin Scheduling

12. Group 896 GPRS Radio Resource Management Proposed Strategy : Access Control

13. Group 896 GPRS Radio Resource Management Proposed Strategy : Resource Assignment

14. Group 896 GPRS Radio Resource Management Proposed Strategy: Scheduling Example: TBF G=3  30% TBF G=6  60% TBF G=1  10%

15. Group 896 GPRS Radio Resource Management Simulation Model- General Assumptions Only downlink traffic in 1 micro-cell, Transmission time step was set to 20ms(1 RLC block) CellHand-over: No Power Control: No Reuse Pattern: 1/3 Cell Radius: 500m Frequency Hopping: Ideal Random Frequency Hopping MobilityUser Distribution: Uniform Distributed User Moving: TU3 Mobility Tracking: No Mobile Phone Always in “active” state Capable of using 1~4 Time Slot

16. Group 896 GPRS Radio Resource Management C/I --- BLER Assumption We designed a network-level simulator which utilized the simulation result of a link-level simulator in reference[3].  TU3 Model  Reuse pattern 1/3  Ideal Frequency Hopping  Fast fading is considered Mean BLER C/I BLER Mapping Function C/I CS_x BLER

17. Group 896 GPRS Radio Resource Management Radio Channel Model (1) Path Loss --- global mean power (2) Slow Fading --- local mean power (3) Fast Fading --- instantaneous power (1) Path loss (2) Slow fading (3) Fast fading

18. Group 896 GPRS Radio Resource Management C/I Generation 1. Mean C/I --- Path Loss and Reuse Pattern Path Loss PDF(r) pdf(r)=k*r, 15m<r<500m

19. Group 896 GPRS Radio Resource Management 0: Investigated Cell 1~6: Co-channel Cell D: Reuse Distance R: Cell Radius r: MS-BTS Distance 1 2 3 4 5 6 0 D R r Reuse Pattern C/I Calculation

20. Group 896 GPRS Radio Resource Management PDF(r) Mean C/I

21. Group 896 GPRS Radio Resource Management 2. C/I per MS --- Shadowing Effect Lognormal Distribution (mean C/I, STD) C/I for Each Mobile

22. Group 896 GPRS Radio Resource Management Coding Scheme Selection CIR 0~5 CS1 CIR 6~9 CS2 CIR 10~16 CS3 CIR 16~30+ CS4 Throughput vs. C/I

23. Group 896 GPRS Radio Resource Management BLER Generation Random Seed< BLER Block Correct Random Seed Random Seed> BLER Block Erroneous C/I CS_x BLER Generation Mapping Table

24. Group 896 GPRS Radio Resource Management Frequency Correlation Function

25. Group 896 GPRS Radio Resource Management Path Loss Shadowing Effect MS ID Distance Generation Mean C/I C/I per MS CS_x Selection BLER Mapping Random Seed Block Correct Or Erroneous pdf(r)

26. Group 896 GPRS Radio Resource Management Traffic Model The definition of a GPRS session in our model: Packet 1Packet 2Packet 3 Packet Call1Packet Call 2Packet Call 3 Session1Session2Session3 Time[ s] Time[ m] Time[ h] Session3

27. Group 896 GPRS Radio Resource Management Traffics: Mixture of GSM Voice calls and GPRS Data sessions Each session/call was randomly assigned to a Mobile Station(MS), where each MS only represent a propagation condition(C/I) corresponding to a certain distance from the BTS.

28. Group 896 GPRS Radio Resource Management Traffic Generation Method

29. Group 896 GPRS Radio Resource Management 1.1 session assigned to 1MS corresponding to a certain radio condition 2.1MS can have multiple sessions based on radio condition 3.All the sessions are in the BSS to be transmitted 4.No Handover 5.Based on one Cell Configuration 6.No uplink simulation 7.Higher layers(TCP/IP, LLC) is not considered 8.ARQ for uplink is not considered Model Assumptions

30. Group 896 GPRS Radio Resource Management ParametersStochastic Process Parameter Values GSM Arrival TimeExponential Holding TimeExponential GPRS Arrival TimeExponential Session SizeGeometricalP=0.5 Object Size = 50 Kbytes Mean Session Size=100Kbytes Parameters List Traffic Load = 5 ~ 29 Erlang GSM Voice: Traffic_Load*70%;GPRS Data: Traffic_Load*30%

31. Group 896 GPRS Radio Resource Management Mapping GPRS session to RLC blocks Session size (bits) GPRS SESSION time CS-x RLC blocks arrival time Mapping to block Because in simulation the transmission was working on RLC block level.

32. Group 896 GPRS Radio Resource Management Simulator Implementation General Structure – Object Oriented Time-Driven and Event-Motivated Simulation Process – main() and traffic_Model Propagation Concerned Object – MS and air_Interface Proposed RRM features VS “ Best Effort ” Proposed RRM Object – BSS, FIFO_Priority and resource_Pool

33. Group 896 GPRS Radio Resource Management General Structure – Object Oriented Access (); TS assignment (); TBF assignment (); Scheduler (); Resource Update (); BSS Traffic Generator ( ); Traffic Model Transmission ( ); Air Interface Update Propagation ( ); Mobile Station Output ( ); Trace Push ( ); Pop ( ); FIFO Priority Add GSM (); Add GPRS (); Release GSM (); Release GPRS (); Scheduled (); Resource Pool

34. Group 896 GPRS Radio Resource Management Time-Driven and Event-Motivated Simulation Process Traffic Model T, Length, MS GSM events Chain T, Length, MS GPRS events Chain Simulation Process T<T_End T_nex t_Frame<T_next_Event? Transmission Scheduling Resource Update Access Timer Update YesNo Output Trace

35. Group 896 GPRS Radio Resource Management Propagation Concerned Object – MS and airInterface MS: Assign C/I, CS, BLER to each mobile station, the process was carried on during the initialization step. Air Interface: Frame[TRX][TS]; During each RLC block transmission, generate a Random “ 0” or “1” representing “correct” or “error” according to the BLER for the MS which occupy current TS/Block.

36. Group 896 GPRS Radio Resource Management Proposed RRM vs “ Best Effort ” Best EffortProposed RRM Access QueuingFIFOGraded FIFO When access queue full and new request come Block the new request Block the last one in queue and push new one into queue SchedulingRound-RobinGraded Round-Robin Resource Allocation When coming GPRS session has to share a TS with existing ones Depending on how many TFI already in a TS Depending on the sum of grades in a TS Voice Pre-emptionYes

37. Group 896 GPRS Radio Resource Management Proposed RRM Object - 1 – BSS, FIFOPriority and resourcePool Resource Pool: store a table for all the TRX-TS. TS data structure Session ID, Grade TS Type 1: GSM Session ID, Grade TS Type 2: GPRS Pool[TRX][TS] TS0TS1TS2TS3TS4TS5TS6 TRX0 TRX1 TRX2 TRX3

38. Group 896 GPRS Radio Resource Management Proposed RRM Object – 2 FIFO_Priority: The GPRS access queue based on grade. Queue length = 7 Session ID, Priority Access Queue Session ID, Priority Priority: High  Low Priority = C/I of each session according to its assigned Mobile Station BSS will always select the first one(highest priority) in queue to assign resources. For two sessions with the same C/I, the new one will be put after the old one(FIFO).

39. Group 896 GPRS Radio Resource Management Proposed RRM Object – 3 BSS: Manage all the “active” sessions. Session ID, Grade GSM Sessions Session ID, Grade GPRS Sessions Session ID, Grade  Access Control (Voice Pre-emption)  TS/TBF assignment  Scheduling  Correctly-received RLC blocks counting  Release finished GSM/GPRS or drop GPRS with max retransmissions.  TFI re-assignment: after every transmission time step, try to raise the TFI numbers for each GPRS session to 4.

40. Group 896 GPRS Radio Resource Management Performance Evaluation Simulator validation Simulations parameters Simulations results Future work

41. Group 896 GPRS Radio Resource Management Simulator Validation RRM strategies: use of output functions and breakpoints to check the behaviour of access queue, choice of TRX/TS, scheduler, reassignment, pre-emption… Propagation /Transmission: comparison between the curve used to map C/I, BLER and throughput and the one obtained Traffic Model: comparison between input parameters and traffic generated (number of sessions per hour, call length, session size…)

42. Group 896 GPRS Radio Resource Management Simulations Parameters Parameter NameVALUE Simulated Duration4 hours Traffic Load (TL)from 5 to 29 "Erlangs" Number of simulations (per TL)15 TRXS4 (32 TSs) Number of signalling channels4 Number of traffic channels28 Maximum number of multiplexed sessions per TS 32 Number of Mobile Nodes30 Multi-slot capacity of Mobile4TSs Mean GSM call duration120s Mean GPRS session size100kB Access queue size7 sessions Maximum number of block retransmission 25

43. Group 896 GPRS Radio Resource Management Simulations Results Throughput per Cell Vs. Traffic Load Throughput per Session Vs. Traffic Load Access Delay Vs. Traffic Load Blocking Rate GPRS Vs. Traffic Load Throughput per Session Vs. C/I

44. Group 896 GPRS Radio Resource Management Throughput per Cell

45. Group 896 GPRS Radio Resource Management Throughput per Session

46. Group 896 GPRS Radio Resource Management Throughput Per Session Vs. C/I

47. Group 896 GPRS Radio Resource Management Mean access delay

48. Group 896 GPRS Radio Resource Management Blocking Rate GPRS

49. Group 896 GPRS Radio Resource Management Future Work Consideration of QoS requirements (traffic model with different types of service) Consideration of Uplink Investigation of higher layers

50. Group 896 GPRS Radio Resource Management Thank You !

51. Group 896 GPRS Radio Resource Management BackUp: for Q&A 1) In traffic generator while calculating the mean GPRS holding time: We use average data speed: 25kbps. Which comes from some simulation output on different traffic load. 2)