IEEE 802.11e Performance Evaluation November 2003 IEEE 802.11e Performance Evaluation Javier del Prado and Patrick Wienert Philips javier.delprado@philips.com Javier del Prado et al. Philips

Outline Goal Overview of Protocols Evaluated Simulation Results November 2003 Outline Goal Overview of Protocols Evaluated Simulation Results Conclusions References Javier del Prado et al. Philips

November 2003 Goal Evaluate the Throughput Performance of IEEE 802.11e MAC protocols as a function of the Frame Size and the Number of Stations Verify that the Usage Scenarios defined in 11-03-802 are implementable Javier del Prado et al. Philips

Overview of Protocols Evaluated November 2003 Overview of Protocols Evaluated Frame exchanges used in the simulation Javier del Prado et al. Philips

Access Point has frames for Station November 2003 DCF Busy Access Point has frames for Station Access Point Beacon t Backoff DIFS Frame 1 Backoff DIFS Frame 2 Ack SIFS Beacon Interval. Typical value = 100 ms Ack SIFS Wireless Station Javier del Prado et al. Philips

(Reduces Backoff Overhead) November 2003 EDCA TXOP Bursting (Reduces Backoff Overhead) t < EDCA TXOP limit Access Point Beacon t Busy Backoff AIFS Frame 1 Frame 2 Ack SIFS Backoff AIFS Ack SIFS Wireless Station Javier del Prado et al. Philips

EDCA with No ACK November 2003 t t Javier del Prado et al. Philips t < EDCA TXOP limit t < EDCA TXOP limit Wireless Station Busy Backoff AIFS Frame 1 Frame 2 Frame 3 Backoff AIFS Frame 1 Frame 2 Frame 3 SIFS SIFS SIFS SIFS t Access Point t Javier del Prado et al. Philips

(Reduces Poll and Ack Overheads) November 2003 HCCA UPLINK TXOP DOWNLINK TXOP Polled TXOP limit Downlink TXOP limit Beacon Piggybacking (Reduces Poll and Ack Overheads) Access Point Busy Poll + Data Ack Ack Data Data PIFS SIFS SIFS PIFS SIFS t SIFS SIFS SIFS Wireless Station Data + Ack Data Ack t Javier del Prado et al. Philips

HCCA with NO ACK November 2003 t t UPLINK TXOP DOWNLINK TXOP Polled TXOP limit Downlink TXOP limit Beacon Access Point Busy Poll + Data Data Data Data PIFS SIFS PIFS t SIFS SIFS SIFS SIFS Wireless Station Data Data Data t Javier del Prado et al. Philips

November 2003 Simulation Results Javier del Prado et al. Philips

BriarLAN 10.0 Simulator Based on OPNET 10.0 PL2 (October 2003) November 2003 BriarLAN 10.0 Simulator Based on OPNET 10.0 PL2 (October 2003) Model Library: 10-24-2003 Functionality Full 802.11 Legacy MAC (DCF + PCF) Most of 802.11e Implementation (Draft 5.0) New Frame Format EDCA EDCA TXOP Bursting Polled TXOP Request Polled TXOP Bursting No ACK Policy DLP Javier del Prado et al. Philips

Throughput vs. Frame Size November 2003 Throughput vs. Frame Size Simulation Scenario: There are two senders and two receivers The senders generate at an infinite rate, L-byte long data frames The frames are not fragmented The channel is error free The beacon interval is 100 ms The underlying PHY layer is 802.11a The Data transmission rate is 54 Mb/s The ACK transmission rate is 24 Mb/s In EDCA simulations AC_VO parameters are used For HCCA The Simple Scheduler defined in 802.11e D5.0 is used CFP is 90 ms During CP, wireless STAs use AC_VO Five frames can be transmitted within a TXOP. Javier del Prado et al. Philips

Throughput vs. Frame Size November 2003 Throughput vs. Frame Size Javier del Prado et al. Philips

Throughput vs. Number of STAs November 2003 Throughput vs. Number of STAs 6 different homogeneous scenarios All stations in each scenario run the same MAC protocol, with same access parameters Load Each STA generates 4 Mb/s of traffic Packet size of 1500 bytes EDCA AC default parameters HCCA Simple scheduler with TXOPs of 5 ms Javier del Prado et al. Philips

Throughput vs. Number of STAs November 2003 Throughput vs. Number of STAs Javier del Prado et al. Philips

Simulation Results for Usage Scenario 1 November 2003 Simulation Results for Usage Scenario 1 Reference: 11-03-802 Javier del Prado et al. Philips

Traffic Generation As defined in 11-03-802-01 November 2003 Traffic Generation As defined in 11-03-802-01 With the Application parameters as shown in the Applications table Added VoD control channel for STA4 STA11 is not simulated – Video Gaming Controller 0.5 Mb/s with 50 bytes packets?  1250 packets/sec? Delay Limit of 4 ms? Javier del Prado et al. Philips

November 2003 Application Set-Up Javier del Prado et al. Philips

November 2003 TCP Parameters Javier del Prado et al. Philips

November 2003 TCP Parameters Javier del Prado et al. Philips

November 2003 WLAN Parameters Javier del Prado et al. Philips

Internet Streaming + MP3 Audio November 2003 EDCA Parameters Default Parameters defined in 802.11e D5.0 Application EDCA AC HDTV AC_VI SDTV Internet File AC_BE VoIP AC_VO Internet Streaming + MP3 Audio Local File Transfer Video Phone We have not tried to optimize EDCA Javier del Prado et al. Philips

HCCA Parameters Used the Simple Scheduler in 802.11e D5.0 November 2003 HCCA Parameters Used the Simple Scheduler in 802.11e D5.0 Polling Interval of 30 ms Minimum Delay Limit of all applications in Scenario 1 We have not tried to optimize Scheduling Algorithm Javier del Prado et al. Philips

PHY Layer 802.11a frame format and IFSs Error free channel November 2003 PHY Layer 802.11a frame format and IFSs Error free channel All STAs transmit data at 216 Mb/s Control and Management Frames at 24 Mb/s Javier del Prado et al. Philips

November 2003 Javier del Prado et al. Philips

DCF Simulation Results November 2003 DCF Simulation Results Wireless LAN data dropped Data is only dropped at the AP The AP has the same priority than other STAs Same probability to access the medium But the AP carries about 50 Mb/s Javier del Prado et al. Philips

EDCA simulation results November 2003 EDCA simulation results Wireless LAN Throughput Wireless LAN Data Dropped Need to calculate PLR from these results + delay curves Javier del Prado et al. Philips

EDCA simulation results November 2003 EDCA simulation results SDTV/HDTV End-to-End Delay Javier del Prado et al. Philips

EDCA simulation results November 2003 EDCA simulation results VoIP End-to-End Delay UPLINK Javier del Prado et al. Philips

EDCA simulation results November 2003 EDCA simulation results VoIP End-to-End Delay DOWNLINK Javier del Prado et al. Philips

EDCA simulation results November 2003 EDCA simulation results TCP Throughput Javier del Prado et al. Philips

HCCA simulation results November 2003 HCCA simulation results Wireless LAN Throughput Wireless LAN Data Dropped Most of data dropped is TCP traffic Javier del Prado et al. Philips

HCCA simulation results November 2003 HCCA simulation results HDTV/SDTV End to End Delay CDF of HDTV/SDTV End to End Delay Javier del Prado et al. Philips

HCCA simulation results November 2003 HCCA simulation results VoIP End-to-End Delay UPLINK Javier del Prado et al. Philips

HCCA simulation results November 2003 HCCA simulation results TCP Throughput Javier del Prado et al. Philips

November 2003 Conclusions Evaluation of the IEEE 802.11e MAC Throughput as a function of frame size and number of stations IEEE 802.11e MAC can achieve high efficiency for large frame sizes Up to 90% Between 55 and 80% for a 1500 bytes long frame HCCA throughput is independent of number of Stations Javier del Prado et al. Philips

Conclusions Simulation of Usage Scenario I November 2003 Conclusions Simulation of Usage Scenario I Both EDCA and HCCA guarantee the required PLR in this simulation scenario There is some data dropped, but given low delay achieved, we can play with the buffer sizes For a 216 Mb/s PHY rate, the network is close to its limit We need to include Channel Errors Do we need to relax the Usage Scenarios definition? Specially for applications with small packet sizes and low delay limit Javier del Prado et al. Philips

November 2003 References [1] Sunghyun Choi, Javier del Prado, Stefan Mangold, and Sai Shankar, “IEEE 802.11e Contention-Based Channel Access (EDCF) Performance Evaluation,” in Proc. IEEE ICC’03, Anchorage, Alaska, USA, May 2003 [2] Javier del Prado, Sunghyun Choi and Sai Shankar "IEEE 802.11e EDCF: a QoS Solution for WLAN"  2nd New York Metro Area Networking Workshop, 2002 [3] Javier del Prado and Sunghyun Choi, “Link Adaptation Strategy for IEEE 802.11 WLAN via Received Signal Strength Measurement,” in Proc. IEEE ICC’03, Anchorage, Alaska, USA, May 2003. [4] S. Mangold, G. Hiertz, B. Walke, “IEEE 802.11e Wireless LAN - Resource Sharing with Contention Based Medium Access,” in PIMRC 2003 [5] S. Mangold, “Analysis of IEEE 802.11e and Application of Game Models for Support of Quality-of-Service in Coexisting Wireless Networks PhD Thesis,” June 2003 [6] S. Mangold, S. Choi, G. Hiertz, O. Klein, B. Walke, “Analysis of IEEE 802.11e for QoS Support in Wireless LANs,” Communications Magazine, Dec 2003 [7] S. Mangold, S. Choi, P. May, O. Klein, G. Hiertz, L. Stibor “IEEE 802.11e Wireless LAN for Quality of Service,” European Wireless 2002 Javier del Prado et al. Philips

November 2003 Backup Slides Javier del Prado et al. Philips

Where are the overheads coming from? November 2003 Where are the overheads coming from? For a 100 bytes packet and 11a PHY Javier del Prado et al. Philips

Where are the overheads coming from? November 2003 Where are the overheads coming from? For a 100 bytes packet and 11a PHY Javier del Prado et al. Philips

Where are the overheads coming from? November 2003 Where are the overheads coming from? For a 1500 bytes packet and 11a PHY Javier del Prado et al. Philips

Where are the overheads coming from? November 2003 Where are the overheads coming from? For a 100 bytes packet and “11n” PHY at 216 Mb/s Javier del Prado et al. Philips

Where are the overheads coming from? November 2003 Where are the overheads coming from? For a 1500 bytes packet and “11n” PHY at 216 Mb/s Javier del Prado et al. Philips

Where are the overheads coming from? November 2003 Where are the overheads coming from? For a 1500 bytes packet and “11n” PHY at 216 Mb/s Javier del Prado et al. Philips