Analysis of Proportional Fair Scheduling in OFDMA Systems with Delay QoS Constraints

1. Analysis of Proportional Fair Scheduling in OFDMA Systems with Delay QoS Constraints SeongwooAhn*, Seokwon Lee**, Kyungho Kim*, and Daesik Hong** *Modem SoC Lab, DMC R/D Center, Samsung Electronics **Information and Telecommunication Lab. (ITL), Department of Electrical and Electronic Eng., Yonsei University

2. Intro. Motivations & Objectives • Delay-QoS Provisioning in OFMDA systems • Impact of delay-QoS constraints on the system performances? • How to increase delay-constrained throughput? • MU div. based throughput increasesacrifices delay-QoS. How to increase both ? Increase of Throughput Increase of Delay QoS Throughput-Delay Tradeoff Objectives of dissertation • Mathematical model for the throughput-delay tradeoff • Efficiently exploitation MU & Freq. diversity with delay-QoScontraint.

3. Preliminary of Effective Capacity Effective Capacity [1] (duality of Effective bandwidth[2]) Effective capacity with Queue length violation Prob. With Delay Constraints (by Little’s Law : Q = λD) Statistical QoS Guarantee QoS triplet with delay-bound violation prob. Part I θ : QoS exponent . It represents QoS level. delay QoS exponent [1] Effective Capacity: A Wireless Link model for Support of Quality of Service - Dapeng Wu, RohitNegi, IEEE TWC, July 2003 [2] Effective Bandwidth in High-Speed Digital Networks - Cheng-Shang, Joy A. Tomas, IEEE JSAC, Aug 1995

4. Part II System Model • System Model • Radio Resource Allocation • M : Degree of Multiplexing (DoM) • N : Degree of Frequency Diversity (DoFD) • Normalized SNR based user selection [6] for Proportional Fairness Service indicator DoM

5. Part II Effect of DoM in Single Ch. Systems (N=1) Per-user EC (for large K (Appox1)) System BW Non-service prob. frame duration determines MU div. gain Mdetermines trade-off of MU div. and Service Prob. M is key parameter on Throughput-Delay Tradeoff

6. Part II Relative Multiplexing Gain • Definition : 1 2 Asθ0, MU div. gain isdominant As θ∞, service prob. isdominant Increase gain as MK Decrease gain as MK

7. Part II Optimal DoM • Proposed scheme (Optimal DoM) • Idea : Use optimal DoM maximizing EC. • Using Approx1 • Integer programming ~ O(K2) • Complex for large K. • Using Relaxation (Approx2) • Convex problem Service prob. = Normalized DoM

8. Part III Effect of DoFD in OFDMA Systems • Frame Structure Q: Increasing DoFD is always helpful to improve the throughput? A: No. DoFD improves only the delay QoS w/o increasing Shannon capacity. MU div. gain Freq. div gain

9. Part II Heterogeneous Case(1) • Assumptions • System concurrently supports a few types of services • Each type of service has the different delay QoS and min throughput.  Reasonable for the practical system • Frame Structure • One Segment for each service type • Segment size is determined base on traffic load by admission controller • Each segment is divided by M*type

10. Part III Simulation results • Heterogeneous Case (a) Video Service (Dmax = 150ms, ε = 10-3) (b) Audio Service (Dmax = 50ms, ε = 10-2)

11. Conclusions • Summary of Contributions Motivation Delay-QoS Provisioning in OFDMA sys. Throughput-Delay Tradeoff Impact factors ? How to improve ? Solution exploit MU div. efficiently Contribution QoS Constraints, DoM, DoFD Optimal DoM • Mathematical Model of Throughput-Delay Tradeoff in OFDMA sys. • Analysis of impact factors on QoS Performance. • Analysis of Optimal DoM in various environments.

12. Thank You ! Q&A