12.07.2015 Views

System Level Modeling and Optimization of the LTE Downlink

System Level Modeling and Optimization of the LTE Downlink

System Level Modeling and Optimization of the LTE Downlink

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

5. Performance Evaluation <strong>of</strong> Fractional Frequency Reuse in <strong>LTE</strong>ness loss), throughput <strong>and</strong> fairness performance has been evaluated over a range <strong>of</strong>number UEs/cell values. Small confidence intervals are ensured by averaging overenough independent different channel realization sets <strong>and</strong> UE positions such thateach plotted point is obtained from averaging at least 500 UE throughput points.The results, shown in Figure 5.15, depict on addition <strong>the</strong> MU-gain analysis resultsfrom Appendix D, so as to compare <strong>the</strong> FFR MU gain results to those <strong>of</strong> differentreuse-1 scheduling strategies. Results indicate that, in order for a static FFR schemesuch as <strong>the</strong> one assumed in this chapter to work, at least 5 UEs per cell are necessary.With less, throughput results do not converge to <strong>the</strong> FFR result. While for <strong>the</strong> roundrobin case, fairness can be consistently increased at no throughput cost, it is alsoclear that <strong>the</strong> same behavior does not hold for PFs. At <strong>the</strong> optimum trade-<strong>of</strong>f point,both cases do <strong>of</strong>fer similar results. While <strong>the</strong> achievable fairness is slightly lowerfor PF than round robin (<strong>the</strong> PF scheduling algorithm pushes aways from extremeresults), throughput is slightly better for PF, as demonstrated in Figure 5.13. BestCQI scheduler results are shown as comparison to <strong>the</strong> maximum achievable multiusergain, which <strong>of</strong> course comes also at <strong>the</strong> expense <strong>of</strong> fairness.Avg. cell throughput [Mbit/s]240220200180160140120100806040Throughput10.9Fairness <strong>of</strong> UE throughput0.80.70.60.50.40.30.20.105 10 15 20 25 30number <strong>of</strong> UEs/cellFairness5 10 15 20 25 30number <strong>of</strong> UEs/cellRound Robin Proportional fair Best CQIRR no cell thr. loss PF no cell thr. lossRR max. fairness PF max. fairnessFigure 5.15: Performance over number <strong>of</strong> UEs per cell for <strong>the</strong> following schedulers: roundrobin, proportional fair, best CQI, FFR-enabled round robin, <strong>and</strong> FFR-enabledproportional fair. The performance <strong>of</strong> <strong>the</strong> <strong>the</strong> FFR-enabled schedulers is evaluatedfor <strong>the</strong> FFR configurations were (i) <strong>the</strong> optimum trade-<strong>of</strong>f between fairness<strong>and</strong> mean throughput is achieved <strong>and</strong> (ii) for <strong>the</strong> case where no mean throughputloss occurs compared to <strong>the</strong> reuse-1 case. Left: cell throughput. Right: UEthroughput fairness. Vertical lines: 95 % confidence intervals.The conclusion <strong>of</strong> this chapter is that, taking into account that PF scheduling isanyway used, due to its increased throughput at <strong>the</strong> same fairness level compared toround robin scheduling, <strong>the</strong> usefulness <strong>of</strong> FFR is limited to allowing a flexible fairness79

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!