210 likes | 414 Views
EDCF TXOP Bursting Simulation Results. Javier del Prado and Sunghyun Choi Philips Research USA Briarcliff Manor, New York sunghyun.choi@philips.com. References. IEEE 802.11e QoS draft D2.0 IEEE 802.11-01/566r3: “ Multiple Frame Exchanges during EDCF TXOP”. Problem Statement.
E N D
EDCF TXOP Bursting Simulation Results Javier del Prado and Sunghyun Choi Philips Research USA Briarcliff Manor, New York sunghyun.choi@philips.com S. Choi, Philips
References • IEEE 802.11e QoS draft D2.0 • IEEE 802.11-01/566r3: “Multiple Frame Exchanges during EDCF TXOP” S. Choi, Philips
Problem Statement • Per 802.11e/D2.0, QSTA cannot transmit multiple MSDUs within an EDCF TXOP • Why not? • As showed in 01/566r3, EDCF TXOP bursting has some advantages while adding minimum complexity S. Choi, Philips
Advantages of EDCF TXOP Bursting • Reduce network overhead: • Without bursting available, a QSTA must backoff after each MSDU transmission • Multiple transmissions using SIFS • Burst Acknowledgement can be used • Bandwidth fairness among the same priority queues, independent from the frame sizes • The HCF polled access is not affected since the EDCF TXOP is limited by the “CP TXOP limit” determined by the HC S. Choi, Philips
Simulation scenario • Fixed data rate of 11 Mbps • 8 QSTAs: 4 voice QSTAs, 4 video QSTAs • Traffic pattern: S. Choi, Philips
Simulation scenario I • EDCF parameters: • Simulations for different TXOP lengths S. Choi, Philips
Simulation results I 2 frames of video per TXOP 3 frames of video per TXOP • Global Throughput No Bursting 3.5 ms 5 ms S. Choi, Philips
Simulation results I • Global Data Dropped No Bursting 3.5 ms 5 ms S. Choi, Philips
Simulation results I • Delay Voice No Bursting 3.5 ms 5 ms S. Choi, Philips
Simulation results I • Delay Video No Bursting 3.5 ms 5 ms S. Choi, Philips
Simulation results I • Throughput video streams No Bursting 3.5 ms 5 ms S. Choi, Philips
Simulation scenario II • EDCF parameters: S. Choi, Philips
Simulation results II 3 frames of video per TXOP • Global Throughput 2 frames of video per TXOP 3.5 ms 5 ms S. Choi, Philips
Simulation results II • Data Dropped 3.5 ms 5 ms S. Choi, Philips
Simulation results II • Delay Voice 3.5 ms 5 ms S. Choi, Philips
Simulation results II • Delay Video 3.5 ms 5 ms S. Choi, Philips
Simulation scenario III • 6 QSTAs, 1 priority • Traffic patterns: • 3 QSTAs for each set of traffic pattern • Network overloaded S. Choi, Philips
Simulation results III • Throughput - No Bursting Video 1 Video 2 S. Choi, Philips
Simulation results III • Throughput - 7 ms TXOP Video 1 Video 2 S. Choi, Philips
Conclusions • EDCF TXOP Bursting increases global throughput when the network is highly loaded • It reduces the global delay • It may increase the delay of streams with low load. This delay can be reduced by adapting the EDCF parameters (as shown in scenario II) • Provides bandwidth fairness among queues with the same priority and different frame sizes (as shown in scenario III) S. Choi, Philips
Conclusions II • In general, we observe a better performance in scenario II in terms of throughput and delay due to a lower probability of collision between QSTAs • The HCF polled access performance should not be affected since the HC controls the duration of the EDCF TXOP • Minimum complexity added optionally as EDCF TXOP bursting will be optional ! S. Choi, Philips