Slide attached.
From: Sindhu Verma <000011381223f2e2-dmarc-request@xxxxxxxxxxxxxxxxx>
Sent: Thursday, July 11, 2024 10:10 AM
To: STDS-802-11-TGBN@xxxxxxxxxxxxxxxxx
Subject: [EXT] [STDS-802-11-TGBN] Discussion on 24/0639r1 "MAC protocol aspects of multi-AP coordination"
Hi all,
Thanks for the discussion on the presentation 24/0639r1 "MAC protocol aspects of multi-AP coordination" during the TGbn Joint teleconference on 20th June.
There were several questions posted in the chat as well as asked online. We have categorized them below into broader topics and embedded detailed responses inline.
Please add any questions that have got missed.
-
Applicability/Extensibility to other types of coordination AP operation
-
This presentation focuses on the per TXOP coordination for C-BF (i.e. M-MIMO coordinated with multiple APs). However, as noted, a similar framework of coordination is applicable for other coordinated multi-AP schemes (e.g., C-SR, CTDMA) as well.
-
Sounding details and overhead:
-
The focus of the presentation is the per-TXOP coordination while providing a high-level overview of the other stages of coordination. The MAC overheads would be largely from the per-TXOP coordination steps.
-
The sounding details will be presented separately. There are multiple options possible such as sequential or joint sounding each with a different MAC overhead. The sounding overhead for C-BF per-se is only due to the acquisition of cross-BSS CSI . The frequency
of sounding is no more than what is needed for general beamforming and legacy DL MU-MIMO
-
In the simulations, the sounding overhead is not accounted for as it would depend on the sounding frequency which is dependent on implementation. Nonetheless, we expect it to be minimal and similar to the overheads seen with DL-MUMO given a typical sounding
frequency of once every several 10s of ms.
-
Per-TXOP coordination steps:
-
Why are responses needed from AP1’s clients to the 1st ICF?
-
Responses are needed to confirm the clients’ participation in the operation of many 11bn features as well as EMLSR. It is also required to confirm that the clients have performed the expected action such as a switch to a certain link or subband or power state
as a result of receiving the ICF. Sending this response also serves to protect the TXOP from hidden nodes
-
The responses also help AP1 identify the clients that will be a part of the CBF transmission
-
Why is feedback necessary from AP2’s clients as a response to the 2nd ICF?
-
For the same reasons as above, i.e. the responses serve to confirm the reception of the ICF at the clients and their expected actions. They also serve to protect the TXOP from hidden nodes and identify the clients to be addressed in the CBF transmission
-
The CBF preamble can benefit from using a common BSS color and TXOP duration to ensure: i) correct decoding of preamble at all listening devices, ii) proper deferral, iii) also helps operation alongside UHR features like NPCA where non-APs within any of the
coordinated APs must not treat the transmissions as OBSS or NPCA opportunity and BSSs different from the 2 coordinated BSSs should be able to treat the CBF transmissions as NPCA opportunity by being able to decode the BSS color
-
Why does the no-CBF throughput stay constant (normalized at 1.00) for any SNR value even as X increases (i.e. interference reduces) while CBF throughput keeps increasing?
-
No-CBF throughput does not change with change in X as long as SNR remains the same because it is based on the initial assumption that the 2 APs hear each other and mutually defer. So, without CBF, they always transmit in a TDM fashion and for a given SNR, throughput
remains constant. Other mechanisms like spatial reuse will have to be applied for the no-throughput CBF to increase
-
Why is there a slight drop in the normalized CBF throughput towards the highest value of SNR for X=0 and X=10?
-
The CBF throughput is a function of what MCSs get used as a part of the CBF operation and there can be some non-linearity in the gains
-
Usage in practical scenarios:
-
Why has full-buffer traffic been used instead of finite-rate traffic?
-
Full-buffer traffic is used often to demonstrate the potential benefits of any feature without making other dependencies like scheduling, link adaptation, start time staggers, etc complicate the results. The improvements seen in full buffer throughput would
be correlated with the performance seen in a congested environment or in the presence of bursty traffic for the duration of the burst.
-
If there is no data to transmit at the other AP, CBF cannot be used
-
Such arguments can be used to counter the benefits of many features like MU-MIMO or OFDMA i.e. they can be used only when other clients with traffic at the same time are present. In most realistic scenarios, this is what happens. Traffic accumulates for multiple
clients and can be transmitted in this fashion. Further, since the pool of candidate clients is larger with two APs versus one, the opportunities for CBF operation could be more than that of single-AP MUMIMO/OFDMA operation.
-
AP1/AP2 decoding frames from each other: This provisioning is required for any coordinated AP features to operate. Keys for Management and Control frames can be negotiated in the stage of AP identification. After all, CBF operation would be possible only among
trusted APs
-
AP1/AP2 decoding frames from the other APs clients: Only beamforming feedback may need to be decoded in this manner. Similar cases may arise for other coordinated AP operations as well.
To unsubscribe from the STDS-802-11-TGBN list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBN&A=1
To unsubscribe from the STDS-802-11-TGBN list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBN&A=1
|