100	5. Evaluation of IEEE 802.11e with the IEEE 802.11a Physical Layer

The fact that QBSS 2 gains from dynamic LA that is applied in QBSS 1 is an undesirable result. There is no motivation to apply spectrum efficient and complicated techniques if the gain from such an effort is shared between coexisting wireless LANs. From the regulatory perspective, radio systems that operate spectrum efficiently must benefit from it. To attract vendors to implement dynamic LA or other radio resource control schemes into their radio systems, other colocated radio systems should not gain equally from its usage.

This problem is known as the “tragedy of commons” in game theory and especially important for radio systems that share unlicensed bands (Salgado-Galicia et al., 1997).

5.2.3.3Throughput Results with Link Adaptation applied in one QBSS and CFBs applied in both QBSSs

Figure 5.27, right, shows the resulting throughput when CFBs are used by both QBSSs. QBSS 1 is capable of applying dynamic LA. Now the throughput of the medium priority streams in QBSS 1 exceeds the throughput of the medium priority streams in QBSS 2. The reason is obvious: after a short transmission of an MSDU, the HC of QBSS 1 is allowed to deliver another MSDU without contending for the access to the channel again, as long as the TXOPlimit is not exceeded (here, TXOPlimit=2.88 ms). Therefore, it is now mainly QBSS 1 that notably improves its performance by applying dynamic LA, compared to the previous scenario. The QBSS 1 improves its performance by efficiently utilizing a given TXOP because of transmitting frames at 54 Mbit/s when possible.

5.2.4Delay Results with CFBs

Figure 5.28 (a) and (b) show the MSDU Delivery delay distributions for both QBSSs in a lightly loaded scenario, i.e., 320 kbit/s for medium and low priority streams, 256 kbit/s for high priority streams. In each figure, the results for one QBSS are shown, where two Complementary Cumulative Distribution Functions (CCDFs) per priority class are given, one for the near station and one for the far station, respectively. It is visible that the LA within QBSS 1 results in considerable shorter minimum MSDU Delivery delays than in QBSS 2.

Due to the higher error probability with the higher PHY modes, retransmissions are more likely in QBSS 1. This is the reason for the higher probability of larger delays in QBSS 1. The near and far stations show different delays in QBSS 1 due to different PHY modes. Figure 5.28 (c) and (d) present the delays when CFBs are applied. It is visible that in a lightly loaded scenario, CFBs have minor impacts on the MSDU Delivery delay.

5.2 Contention Free Bursts

101

(a) QBSS 1, no CFBs. LA only in this QBSS.

(c) QBSS 1, CFBs in both QBSSs. LA in this QBSS.

(b) QBSS 2, no CFBs. LA in other QBSS 1.

(d) QBSS 2, CFBs in both QBSSs. LA in other QBSS 1.

Figure 5.28: MSDU Delivery delays. Left: QBSS 1, Right: QBSS 2.

As before, QBSS 1 always shows smaller delays than QBSS 2, as the transmission times in QBSS 1 are reduced with the higher PHY modes. Figure 5.28 (d) indicates that QBSS 1 fills its TXOPs often up to the TXOPlimit of 2.88 ms, which is the reason for the shape of the curve of the high priority streams within QBSS 2.

5.2.5Conclusion

The concept of CFBs is an attractive element of IEEE 802.11e in terms of spectrum efficiency, and economy. In overlapping QBSS coexistence scenarios, a wireless LAN takes advantage of applying dynamic link adaptation when CFBs are used. A wireless LAN that uses CFBs can improve its performance compared to other wireless LANs that operate without CFBs. With CFBs, future wireless LANs will apply dynamic link adaptation in order to achieve an higher throughput. Without CFBs, future wireless LANs will not necessarily apply dynamic link adaptation. Without CFBs, coexisting wireless LANs achieve the same throughput results. The use of the CFB mechanism motivates for the application of link adaptation, which as a result increases the spectrum efficiency of radio systems in the unlicensed 5 GHz band.