1. Dynamic Bandwidth Allocation with Fair Scheduling For WCDMA Systems Liang Xu, Xumin Shen, and Jon W. Mark University of Waterloo published in IEEE Wireless Communications Magazine, vol. 9, no. 2, April 2002 2005/010/12 Presented by Hojin Lee(lumiere@mmlab.snu.ac.kr)lumiere@mmlab.snu.ac.kr

2. 2 Contents System Model Ideal Scheduler: GPS Related Work  SCDMA – Scheduled CDMA Proposed Scheduler  CDGPS Performance Evaluation Conclusion & Comment

3. 3 System Model CDMA – code division multiple access  Chip rate  Multi-code  Spread factor  Power control – close loop FDD – frequency division duplex Each Mobile user is assigned to at least one dedicated channel, and shares the random access channel (uplink)  Dedicated channel – date  Random access channel – signaling, short message Total uplink capacity is limited by intracell and intercell multiple access interference

4. 4 CDMA: ex)

5. 5 Ideal Scheduler: GPS GPS – Generalized Processor Sharing  Max-min fair allocation Maximizes the minimum share of a source whose demand is not fully satisfied  Intuitively, GPS serves packets as if they are in separate logical queues, visiting each nonempty queue in turn and serving an infinitesimally small amount of data.  In any finite time interval, it can visit every logical queue at least once.  GPS can’t be implemented Packet-by-packet GPS  Finish tag, virtual time, start tag capacity

6. 6 SCDMA – Scheduled CDMA At the beginning of each time slot, BS chooses as many MSs with the smallest finish tags as permitted by the system power index budget  => excessive computational complexity  => signaling overhead

7. 7 CDGPS – Code Division GPS : total amount of backlogged traffic of flow i at the end of time slot k without service   : the end time of slot k-1  : the backlogged traffic at time  : estimated traffic arrival rate of flow i during slot k If == 0, = 0 If > 0, =, If < CT  The remaining resource is distributed to users who expected more than their guaranteed service

8. 8 Performance Evaluation (1/2) C = 2Mbps Perfect power control T: 10 ms 4 homogenous best effort packet data flow L: 5120 bits Poisson process, shaped by leaky bucket regulator with = 20L and = C/4

9. 9 Performance Evaluation (2/2) 10 voice flows  on-off model - on: 40%, CBR 16 kbps, 80 bits 3 video flows  VBR  8-state MMPP model - Average duration in each state: 40 ms 4 best effort data flow  Poisson process - 2560 bits, 256 kbps

10. 10 Conclusion & Comments SCDMA => complexity, signaling o/h CDGPS => simple, use idle network resources No comparison CDGPS with SCDMA Simulation environment is not clearly explained Power control is not trivial Soft Capacity => intercell interference