Downlink Scheduling With Economic Considerations to Future Wireless Networks

1. Downlink Scheduling With Economic Considerations to Future Wireless Networks Bader Al-Manthari, Nidal Nasser, and Hossam Hassanein IEEE Transactions on Vehicular Technology, Vol.58, No.2, FEBRUARY 2009 報告人：李宗穎

2. Outline • Introduction & Related Work • System and Packet Scheduler Model • Centralized Downlink Packet Scheduler • Performance Evaluation • Conclusion

3. Introduction (1/2) • Future wireless cellular systems (HSDPA : high-speed downlink packet access) offer high data rates that are beyond the capabilities of 3G systems • A key component of radio-resource management is packet scheduling, which is responsible for distributing the shared radio resources among the mobile users

4. Introduction (2/2) • The packet scheduling scheme should track the instantaneous channel conditions of the connections and select for transmission those that are experiencing good channel conditions to maximize system capacity

5. Research Goal • This paper design CDPS (centralized downlink packet scheduler) to balance between the requirements of connections (throughput & fairness) and the requirements of service providers (revenues)

6. Related Work • maximum carrier-to-interference ratio (Max CIR) [5] • Max CIR tends to maximize the system’s capacity by serving the connections with the best channel quality • proportional fairness (PF) [6] • PF tries to increase the degree of fairness among connections by selecting those with the largest relative channel quality [5] S. Borst, “User-level performance of channel-aware scheduling schemes in wireless data networks,” in Proc. IEEE Conf. Comput. Commun. INFOCOM, Mar. 2003, vol. 1, pp. 321–331. [6] A. Jalali, R. Padovani, and R. Pankaj, “Data throughput of CDMA-HDR ahigh efficiency-high daterate personal communication wireless system,”in Proc. IEEE VTC, May 2000, pp. 1854–1858.

7. Contributions • This paper propose a novel centralized downlink packet scheduler (CDPS) scheme • CDPS tries to balance between the user connection’s preferences (as perceived by the service provider) and the fairness by formulating an optimization problem that can be solved in real time • the service provider can choose the degree of fairness of the CDPS • CDPS can be configured to reduce to the Max CIR and PF schemes

8. System and Packet Scheduler Model • Assumption • Only one connection is scheduled for transmission at each frame • Scheduling scheme will equally work if more than one connection is scheduled • These PDUs are stored in the transmission queue of the corresponding connection in a first-in–first-out fashion

9. Packet scheduler model • Each connection regularly informs the base station • the size of the transport block that the base station should send to the connection • number of simultaneous channel codes, and the type of modulation and coding schemes

10. Centralized Downlink Packet Scheduler • The proposed scheme (CDPS) employs practical economic models through the use of utility and opportunity cost functions • First, outlining the general formulation of CDPS • Second, provide a definition for a possible utility function, an opportunity cost function, and a fairness measure • Finally, mathematically show that our defined utility function for CDPS reduces to the Max CIR and PF scheduling schemes

11. General formulation of CDPS • Xi1(t), . . . , Xim−1(t) are the chosen quantitative measures (ex: data rate, average delay…) • Xim(t) is a fairness measure that represents how fair the scheduling scheme is to the user connection • OCi(t) is the opportunity cost of serving connection i at time t • K is a predefined constant value

12. Opportunity cost function • opportunity cost is how much data rate the system would compromise if connection i is selected for transmission given that there is a connection j with a higher current data rate • Ri(t) is the current data rate for connection i at time t, which depends on its channel condition • maxj Rj(t) is the maximum current data rate of all connections at time t

13. Cobb–Douglas functional form of production functions • Y = ALαKβ • Y = total production • L = labor input • K = capital input • A = total factor productivity • α and β These values are constants determined by available technology • α + β = 1 (constant) • α + β < 1 (decreasing) • α + β > 1 (increasing)

14. Cobb-Douglas Utility Function • Assuming m = 2 in our formulation of CDPS, the Cobb–Douglas utility function is expressed as • Xi1 be any performance metric that the service provider wants to optimize • Xi2 be a fairness measure that increases as the connection’s or system’s perception of fairness increases • c and d is constant value

15. Definition of Xi1(t) and Xi2(t) • The utility of connection i being served increases as Ri(t) increases • γiis used to control the shape of Xi2(t) • αi(t) = Si(t)/ (maxj Sj(t)) • Si(t) is the average throughput for connection i up to time t • maxj Sj(t) is the maximum average throughput achieved among all connections up to time t

16. αi(t) and γi in Xi2(t) • the larger the value of γi, the higher the rate of decrease in Xi2(t) (dynamically changed as needed by the service provider) • the scheduling scheme to be fairer to the connections with low αvalues (i.e., low average throughputs compared with connections with high average throughputs)

17. CDPS decision rule • CDPS will find the connection that would maximize the following objective function: • a solution can be found by choosing connection i for transmission such that

18. Properties of CDPS • Efficiency • CDPS makes efficient use of the bandwidth by relatively favoring connections with good channel conditions • Fairness • CDPS also consider average throughputs compared with the maximum average throughput • User satisfaction • using both the instantaneous channel condition and the user’s connection relative fairness • Flexibility • flexibility to choose the degree of fairness and thereby control the capacity–fairness tradeoff and effect the obtained revenues

19. Flexibility of CDPS • Max CIR • If K is set to 0, then the CDPS reduces to the Max CIR scheme • PF schemes • If c is set to 0, d is set to 1, then the CDPS reduces to the PF scheme

20. Using NS2 + Enhanced UMTS Radio Access Network Extensions Simulation Model Each connection sends a request for one FTP file User Download FTP data (50Mb)

21. Throughput Cell throughput with different values of K Cell throughput 25 user connections

22. User Satisfaction User satisfaction with minimum throughput of 128 kb/s with different valuesof K

23. Conclusion • CDPS scheme for future wireless cellular systems that is based on a utility function to represent the satisfactions of the mobile users as perceived by the service provider