DL MU-MIMO ack protocol Month Year doc.: IEEE 802.11-yy/xxxxr0 DL MU-MIMO ack protocol Date: 2009-11-16 Authors: John Doe, Some Company
Overview Analyze two ack protocol options for DL MU-MIMO: Polled acks Scheduled acks Look at robustness, efficiency and complexity
Polled Acks One user responds immediately to implicit BAR Other users wait to be polled with BAR frame Simple: No change to existing protocol Permitted under existing 11n channel access rules
Scheduled Acks All users respond to implicit BAR Each user has a different offset for response to avoid collision Need to add ‘offset’ field to Data MPDUs Slightly more efficient than polled acks Probably requires NAV protection: Sequence gaps are possible (missing BA) Third party STA may access the medium and collide with scheduled ack
Efficiency gain of Scheduled Acks over Polled Acks TXOP duration 2 STA 3 STA 4 STA 1.5 ms 3.5% 6.9% 10.4% 3 ms 1.7% 5.2% Assuming 52 us overhead for SIFS+BAR (12 Mbps 11a format BAR)
Error recovery with polled acks Normal error recovery rules apply with polled ack Sequence is fully under AP control
NAV Protection for Scheduled Acks Need NAV to protect against sequence gaps (missing BA) L-SIG TXOP protection: has EIFS problem with legacy STAs Prefer RTS/CTS RTS/CTS adds overhead, but is beneficial with long TXOP and on a busy network RTS/CTS acts as a collision detect mechanism: in case of collision, entire TXOP (3ms) is not lost
Conclusion Polled ack mechanism is supported under existing 11n rules Should be supported as baseline mechanism Proposed enhancements should be compared to this technique Scheduled acks has some efficiency gain over polled ack But slightly more complex Protection overhead and benefit applies to both schemes