1 / 25

AP Service Load: Improved Definition

July 2004. doc: IEEE 802.11-04/550r1. AP Service Load: Improved Definition. Joe Kwak InterDigital. Prior MIB definition for AP Service Load.

maxine
Download Presentation

AP Service Load: Improved Definition

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. July 2004 doc: IEEE 802.11-04/550r1 AP Service Load: Improved Definition Joe Kwak InterDigital Joe Kwak, InterDigital

  2. Prior MIB definition for AP Service Load • "This counter shall be a scalar indication of the relative level of service loading at an AP. A low value shall indicate more available service capacity than a higher value. The value 0 shall indicate that this AP is not currently serving any STA. The value 128 shall indicate that the AP is operating at its target service loading. The value 255 shall indicate no additional AP service capacity is available or that this STA is not an AP." Joe Kwak, InterDigital

  3. New Metric Definition • The Medium Access Delay (MAD) metric is defined as the average delay incurred from the time that any DCF packet is ready for transmission (i.e. begins CSMA/CA access) to the actual packet transmission start time. • Since DCF packets are lower priority than PCF or HCF packets, the DCF access delay values are sensitive to all PCF, HCF and DCF channel loads. • While channel is busy for PCF or HCF, DCF backoff counting is suspended while access delay timing continues. Joe Kwak, InterDigital

  4. MAD is sensitive to DCF, PCF and HCF Loads Each beacon interval example yields same MAD result 3msec Access Delay DCF 1. DCF with 20 users at rate A 50% 3msec Access Delay 2. PCF DCF DCF with 10 users at rate A 75% 3msec Access Delay 3. HCF DCF DCF with 5 users at rate A Joe Kwak, InterDigital

  5. MAD Timing Normal Case • Medium Access Delay is measured by access timer which starts when DCF packet begins CSMA/CA access and stops when packet begins transmission on wireless medium. • In Annex C behavior charts for normal burst transmission, the access timer is controlled by two signals into/out of the Protocol_Control_AP process for DCF packets. • PduRequest signal into process starts access timer • TxRequest signal out of process stops access timer. • Only one packet at a time is processed for each control function; PduRequest results in successful transmission or discard (PduConfirm with retrylimit parameter after max retransmissions). Joe Kwak, InterDigital

  6. MAD Timing in Packet Transmission PduRequest to Protocol Control Process PduRequest for next Packet CSMA/CA TxRequest to Transmission Process Optional ACK Packet Transmission Time Transmit Access Delay == MAD, optionally contains RTS/CTS. MAD Timer or timestamps used to measure CSMA/CA duration. Joe Kwak, InterDigital

  7. PduRequest MAD Timing Examples PduRequest Starts Timer TxRequest Stops Timer PduConfirm MAD Timer TxRequest B1 A B2 C C TxOut(Pdu) B3 RxIn(ACK) Normal PDU with Retransmission 3 Fragment PDU Transmission Normal PDU Transmission Joe Kwak, InterDigital

  8. MAD Timing Triggers, Normal PduRequest starts timer, TxRequest stops timer Joe Kwak, InterDigital

  9. MAD Timing for Frag or Retransmit Case • For fragmented Pdu, timer is stopped upon TxRequest of first fragment, then restarted upon initial backoff to begin CSMA access for second fragment within process Tx_Coordination_AP. • For retransmit case, access timer is stopped upon TxRequest, then restarted upon NACK detection within process Tx_Coordination_AP. • Note that PCF and HCF packets may be concurrently processed, but PduRequest/Confirm signals are unambiguously labelled with the transmitted fsdu. • Note also that DCF backoff downcount timing is suspended while DCF detects CCA busy (or NAV set). Joe Kwak, InterDigital

  10. MAD Timing Restart Trigger for Fragment Start MAD Timer Loop reentry for next fragment transmission Restart MAD Timer Joe Kwak, InterDigital

  11. MAD Timing Restart Trigger on Retransmit Restart MAD Timer Joe Kwak, InterDigital

  12. MATLAB Simulation Setup • The simulator models a single BSS with one AP and multiple associated STAs. • Traffic modeling: • Packets arrive for transmission at each STA according a Poisson process. • New Packets are queued for transmission at the AP every time a successful packet is received at the AP from a transmitting STA (i.e. the AP responds to the STAs transmission). The number of packets that are transmitted by the AP per received packet is equal to asymmetry_factor. For example, asymmetry_factor=1 corresponds to symmetric upstream/downstream traffic. When asymmetry_factor=2, there are 2 downstream packets for every upstream packet. • Different offered loads are achieved by varying the alpha_STA, N_STA and asymmetry_factor parameters. Joe Kwak, InterDigital

  13. MATLAB Simulation Setup (cont.) • The 802.11 DCF access method is modeled, i.e. carrier sense (deferral) and collision avoidance (random back-off procedure) are modeled. • All STAs "hear" each other’s transmissions, i.e. no hidden terminals. • Transmission errors due to collisions are modeled. The effects of collisions are accurately modeled by doubling the size of the Contention Window (CW) following transmission errors (i.e. collisions), up to the maximum allowed CW. The CW is reset to the minimum CW upon successful transmission (i.e. without collision). • All collision-free transmissions have sufficient SNR (i.e. perfect channel) to be error free. Joe Kwak, InterDigital

  14. MATLAB Simulation Setup (cont.) • Packets are dropped once the maximum number of retransmissions is attained (default value of 4 is assumed), after which point CW is reset to CWmin. • When a packet is transmitted, the PLCP header, the transmission of the MPDU, the SIFS wait-time (waiting for ACK), the PLCP header of the ACK and the transmission of the ACK MPDU are modeled. • The following metrics are logged individually for the AP and each STA during a simulation run: • Offered load (number of packets and bits/second) • Throughput (number of packets and bits/second) • Medium Access Delay (MAD, in seconds per packet) • MAC Transmit Delay ( MTD, in seconds per packet) where MTD= queueing delay + MAD Joe Kwak, InterDigital

  15. Simulation Results 1 Joe Kwak, InterDigital

  16. Simulation Results 2 Joe Kwak, InterDigital

  17. Simulation Results 3 Joe Kwak, InterDigital

  18. Simulation Results 4 Joe Kwak, InterDigital

  19. Simulation Results 5 Joe Kwak, InterDigital

  20. TGe Load Metric: QBSS Load • Actually provides three elements all relating to load: • Station Count Field, unsigned 16 bit integer indicating total number of STAs associated • Channel Utilization, 8 bit percentage of time the QBSS detects the medium as busy • Available Admission Capacity, 16 bit integer representing amount of medium time (32usec units) available via explicit admission control Joe Kwak, InterDigital

  21. Problems with TGe Metrics • Three metric components make comparative loading evaluations difficult/impossible. • No agreed way to combine the three variables into a summary metric for comparison purposes. • Simulation results show how AP Service Load metric relates to Channel Utilization, number of STAs associated and traffic asymmetry. • New AP Service load metric permits AP loads to be compared between APs operating under different conditions. Joe Kwak, InterDigital

  22. New BSS Load IE • Similar to TGe QBSS Load but modified for Radio Measurement capable APs • Three component elements: • Station Count Field, same as TGe • Channel Utilization, same as TGe • New AP Service Load • Included in Beacons and Probe Response, like TGe • Conditional inclusion of Station Count and Channel Utilization in Beacon and Probe Response to prevent redundant information • Included as new Statistics Group in Statistics Report to provide upper layer MIB interface and radio interface Joe Kwak, InterDigital

  23. New AP Service Load Properties • Quantized to 8 bits like RSSI & RCPI • Based on MAD measurements in AP downlink during contention periods. • logarithmic scaling over meaningful range • Scaled so min value 1 represents 50usec (DIFS), and max value 253 represents 5.5msec covering a 20.4 db range • Special values: • 0 = All capacity available (no STAs associated) • 254 = no capacity available (not accepting new associations) • 255 = AP Service Load not available • AP measures MAD over thirty second window; accuracy of +/-200 usec specified with minimum of 200 packets in average • Combines AP loading effects of #STAs, Channel Utilization (PCF, HCF and DCF) and traffic assymetry into SINGLE metric to compare BSS relative loading. Joe Kwak, InterDigital

  24. New AP Service Load Definition • "The AP Service Load shall be a scalar indication of the relative level of service loading at an AP. A low value shall indicate more available service capacity than a higher value. The value 0 shall indicate that this AP is not currently serving any STA. The values between 0 and 254 shall be a logarithmically scaled representation of the average medium access delay for DCF transmitted packets measured from the time the DCF packet is ready for transmission (i.e. begins CSMA/CA access) until the actual packet transmission start time. Value 1 represents 50 usec, value 253 represents 5.5 msec or any delay greater than 5.5 ms. The value 254 shall indicate no additional AP service capacity is available. The value 255 shall indicate that the AP Service Load is not available. The AP shall measure and average the medium access delay for all transmit packets using DCF mechanism over a continuous thirty second measurement window. The accuracy for the average medium access delay shall be +/- 200 usec or better when averaged over at least 200 packets.” Joe Kwak, InterDigital

  25. Motion for Improved Normative Text • Move to instruct the editor to incorporate text from document 11-04-0632-01-000k-BSS_Service_Load.doc into next TGk draft specification document • Moved by Joe Kwak • Seconded by: _______________ • Vote YEA _______ • Vote NEA _______ • ABSTAIN _______ • Vote Passes/Fails at ___% Joe Kwak, InterDigital

More Related