Hi Minyoung,
Please find my response inline.
Regards
Guogang Huang
发件人: Minyoung Park [mailto:mpark.ieee@xxxxxxxxx]
发送时间: 2022年9月29日
13:10
收件人: huangguogang <huangguogang1@xxxxxxxxxx>
抄送: STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
主题: Re: [STDS-802-11-TGBE] CR doc 1381r1 - Beacon-A frame
Hi Guogang,
Good to see that you agree that the consensus is to allow T2L mapping for both DL and UL. How it is used will be up to implementers.
I don't see a clear answer to my last question. How does information in an A-control field tell which link should be used to get a BU with a TID for a non-AP MLD? The A-Control field should be in a unicast frame transmitted
to the non-AP MLD but the non-AP MLD doesn't know which link to get that frame in the first place?
[Guogang Huang]The newly defined MLTI A-control doesn’t coupled with the T2L mapping. The AP MLD can recommend any link to retrieve buffered BUs. Specifically, there are two cases:
Case 1. The non-AP MLD has at least one link (e.g. Link 1) in the active mode. The AP MLD can directly send a frame with the MLTI A-control field on this active link (e.g. link 1) to wake up these
links to get buffered BUs.
Case 2. All affiliated STAs of a non-AP MLD are in PS mode. In this case, the AP MLD shall set the corresponding bit to 1 in the TIM element to inform this non-AP MLD to retrieve buffered BUs. The
non-AP MLD can issue a PS-poll on any link. If the AP MLD want to change the link to delivery buffered BUs, it can send a frame with a MLTI S-control field.
Hi Minyoung,
Thanks for your response. Please find my response inline.
Regards
Guogang Huang
发件人:
Minyoung Park [mailto:mpark.ieee@xxxxxxxxx]
发送时间: 2022年9月29日
1:54
收件人: huangguogang <huangguogang1@xxxxxxxxxx>
抄送:
STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
主题: Re: [STDS-802-11-TGBE] CR doc 1381r1 - Beacon-A frame
Thanks for sharing your opinion.
Isn't the group's consensus to allow T2L mapping for both uplink and downlink?
[Guogang Huang]Yes, the current T2L mapping can be negotiated for both UL and DL. But
my point is, even the T2L mapping negotiation can support both UL and DL, the real implementation case is it’s the best choice to apply the default mapping for the downlink transmission.
Why is T2L mapping only useful for UL and not for DL?
[Guogang Huang]Because the AP MLD as the manager know the traffic load on each link. The
best choice is to let the AP MLD dynamically schedule the downlink transmission based on the real-time traffic load.
When T2L mapping for DL is set up and not all TIDs are mapped to all enabled links, how does a non-AP MLD know which link to use to get buffered BUs with a TID
that is not mapped to all enabled links?
[Guogang Huang] I had explained the downlink T2L mapping is not very useful. Furthermore,
the downlink T2L mapping is not good for the non-AP MLD’s power saving. For example, TID #n are only mapped to link 2, affiliated STA 2 is in PS mode and affiliated STA 1 is in active mode. If there are buffered BUs with TID #n, according to the current draft,
affiliated STAs shall wake up to retrieve buffered BUs. But in this case, especially the traffic load on link 1 is low, we should allow the non-AP MLD can retrieve buffered BUs on link 1.
Even when T2L mapping for DL is set up and not all TIDs are mapped to all enabled links,
the AP MLD still can inform a non-AP MLD know which link to use to get buffered BUs by defining a MLTI A-control field.
Hi all,
Sorry to interrupt your discussion. I would like share my opinion on the MLTI element.
I don’t think a broadcast Beacon-A frame is a good direction.
Based on the current draft text, the MLTI element is coupled with the T2L mapping. But
this is not good for the non-AP MLD’s power saving. Because In my opinion, the T2L mapping is only useful for the uplink transmission. The T2L mapping is useless for the downlink transmission and it may cause the unnecessary wakeup. Furthermore, the concept
of the downlink T2L mapping conflicts with the data cross-link transmission. In order to fully exploit all enabled links, the best way is the AP MLD can dynamically schedule the downlink transmission among all enabled links by defining a MLTI A-control field.
Hence, I suggest if the Beacon size is larger than a threshold, the MLTI element shall
not include within the Beacon frame. The AP MLD can use the defined MLTI A-control to dynamically schedule the downlink transmission based on the current traffic load.
Regards
Guogang Huang
发件人:
Minyoung Park [mailto:mpark.ieee@xxxxxxxxx]
发送时间: 2022年9月28日
7:24
收件人: STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
主题: Re: [STDS-802-11-TGBE] CR doc 1381r1 - Beacon-A frame
Hi Abhi, Gaurav, and Pooya,
Thanks for sharing your experience dealing with the issue.
I also agree with the issue for a legacy STA dealing with the Beacon size larger than 1.5-2 kbytes depending on vendors.
Agreed with Gaurav. It is actually a problem in the field.
Hi Jay,
We encounter wireless clients from different vendors in the field that expect the Beacon frame to be less than a certain size (thresholds vary from 1508 bytes to ~1800 bytes and some having 2300 bytes). It is not a problem with just a single implementation,
but is a pretty widespread issue. If you see the transmitted Beacon size from 802.11ax capable tri-band (2.4/5/6) APs and with MBSSID + RNR enabled for few Virtual-APs per band, we quickly start to run into these thresholds. Now add to this 802.11be with its
MLE, per-STA profiles, ML Traffic Indication, etc. and the problem with Beacon size is pretty glaring.
We need a solution which not only fixes this problem now, but also is a flexible enough
framework to be extended in the future amendments. I understand that this should not have been the problem in the first place, but we are here, and now we have to fix it without affecting legacy clients. To that end, I think we should move forward with
the direction proposed by Minyoung.
With Regards,
Gaurav Patwardhan
(Hewlett Packard Enterprise)
Hi Abhi,
I consulted some STA vendors, but I don’t hear any vendor mentioned the Beacon parsing issue. I’m not sure whether it’s a general issue in WIFI industry or a corner
issue in some specific vendor’s product.
As you mentioned, it’s a legacy STA issue. Suppose such product is already in the market. Could you provide more evidence on this, like product list and test data?
Why has such special design, due to no enough buffer for the Beacon frame?
Anyway, hope you can provide more information on this so that I can support you.
Thanks
Best Regards
Jay Yang
Hi Jay,
As Minyoung explained, there are legacy STAs from various vendors out in the field that expect the Beacon frame to be within a certain size. These devices fail to
parse the Beacon frame if the size of the frame exceeds a certain value (the exact value varies from 1.5k bytes to 2k bytes depending on the vendor and the product). The contents of the frame doesn’t matter, it is the overall size. When a legacy device fails
to correctly parse the Beacon frame, it starts misbehaving and that brings down the overall system performance.
This is an industry wide issue and is affecting Wi-Fi interop.
As we all are aware, the size of ML Traffic IE can become very large and will easily lead to the size of the Beacon frame exceeding the threshold for many legacy
devices. Let’s not hold a short term view of the problem by addressing it with a patchwork solution. Moving ML Traffic IE out of the beacon is the right direction for the Wi-Fi industry.
Regards,
Abhi
WARNING:
This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.
Maybe George, Abhi, and Brian who submitted comments can elaborate more on the issue.
My understanding is that when a beacon frame becomes too big, legacy STAs have a problem processing that large beacon frame.
WM usage will be similar (maybe additional SIFS and MAC header).
Hi Minyoung,
Your explanation makes me confused.
The legacy STA can’t decode the MLTI element and will discard it naturally, why you thought it’s a problem?
And your proposal will cause extra channel resource wasting issue, seems you don’t object with this point.
Thanks
Best Regards
Jay Yang
From: Minyoung Park <mpark.ieee@xxxxxxxxx>
Sent: 2022年9月26日
14:25
To: Yang, Zhijie (NSB - CN/Shanghai) <zhijie.yang@xxxxxxxxxxxxxxx>
Cc: STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-11-TGBE] CR doc 1381r1 - Beacon-A frame
Hi Jay,
Please see the following comment. The beacon bloating issue is a problem that a legacy STA having a problem with dealing with a large beacon frame and not necessarily
a beacon overhead issue.
|
10386
|
GEORGE CHERIAN
|
9.4.2.315
|
0.00
|
Remove the Multi-Link Traffic Indication element from the beacon, since it can cause beacon bloating, that can affect legacy clients
|
On the second question, the Beacon-A frame is also a group addressed frame.
Hi Minyoung,
I have a general question on your proposal. When we talk the Beacon bloating issue, it equals to the big size of Beacon frame cost too much channel resource. I wonder
how to save the channel resource based on your proposal that splits one Beacon to two Beacons?
Let’s assume TXOP (beacon) = beacon/ data rate; your proposal looks like: TXOP(beacon-1)/data rate + SIFS + TXOP(A-beacon)/data rate. Am I right?
Second, the group addressed BU shall be sent after DTIM beacon immediately according to the baseline rule, and your proposal makes such rule changed, how to compatible
with legacy STA? e.g. Legacy STAs may see the AID0 is set in the partial Bit map in the DTIM beacon, and plan to receive the group addressed BU. But they see a unknown A-Beacon follows the DTIM beacon, the legacy STA will be confused.
Thanks
Best Regards
Jay Yang
From: Minyoung Park <mpark.ieee@xxxxxxxxx>
Sent: 2022年9月24日
2:57
To: STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-11-TGBE] CR doc 1381r1 - Beacon-A frame
Hello all,
Thanks for the feedback on doc 1381r1.
Most of the comments so far are related to further optimization on the size of the MLTI (multi-link traffic indication) element. Although we can spend time and work
on optimizing the size of the MLTI element for many different scenarios, once the MLTI element is moved to a new Beacon-A frame the beacon bloating issue for legacy STAs is resolved and the size of the MLTI element becomes less of a problem. Instead, IMO,
simplifying the MLTI element processing might be a more important aspect to consider. With this in mind, reusing the structure of the Link Recommendation frame that includes the AID Bitmap and the MLTI element seems to be a natural choice.
I've uploaded 1381r2 reflecting the changes on D2.2.
Please see below responses as well:
-
Rojan - although I agree that your proposed approach will reduce overhead of the AID bitmap when there are many zeros in the AID bitmap, this requires using information across TIM
in the Beacon and modified AID Bitmap and MLTI elements in the Beacon-A frame.
-
Vishnu - Since the AID Bitmap element and the MLTI element are in the same frame, whether to have the AID bitmap information in the MLTI element is not a critical part of the design.
The Link Recommendation frame includes the AID Bitmap and the MLTI element so reusing the same structure might be better for simplicity.
-
Xiangxin - The Link Recommendation frame includes the AID Bitmap and the MLTI element so reusing the same structure might be better for simplicity.
-
Shawn - On item1, I revised the resolution for CID 13855. The condition is now updated as "...and the AP MLD has buffered BU(s) (#13855)with
TID(s) that are not mapped to all the enabled links for the non-AP MLD(s)." On item 2, indicating all the links that can be used to retrieve BU with the mapped TID(s) makes
more sense to me.
-
Yunbo - Since we resolved the beacon bloating issue with the Beacon-A frame and removed unnecessary overheads (bitmaps associated with legacy STAs or non-AP MLDs with default mapping)
from the MLTI element, further optimizing the size of the MLTI element seems to be not a critical issue.
Hi Minyoung,
Thanks for your presentation. I have below two comments, would you please let me know your opinions.
1)
Do you consider to keep the AID offset subfield? In that case, it will bring several benefits. Firstly, single frame structure of ML traffic indication element
can be used in Beacon and Beacon-A frame. Secondly, multiple ML traffic indication elements can be carried in same Beacon or Beacon-A frame. Think about a use case that AP MLD allocate segment A of AIDs to non-AP MLDs with two links, and allocate segment B
of AIDs to non-AP MLDs with three link. So that the AP could carry two ML traffic indication elements and the bitmap lengths in two element are 2 and 3 respectively. So it could potentially save the overhead.
2)
Link ID offset subfield could help to reduce the signaling overhead, and a link ID bitmap also could achieve same purpose. In some cases that the link IDs are
not contiguous, a link ID bitmap could do better. If a link ID bitmap is used here, we don’t need to force
that each AP MLD allocate the link ID contiguous, which may release Abhi’s concern.
Regards,
Yunbo
发件人:
Minyoung Park <mpark.ieee@xxxxxxxxx>
发送时间:
2022年9月16日
8:02
收件人:
STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
主题:
Re: [STDS-802-11-TGBE] CR doc 1381r1 - Beacon-A frame
Hi all,
I'm starting an email thread on the Beacon-A frame proposal in CR doc 1381r1. Please post your questions in the thread. I answered the questions posted on the chat
and during the meeting.
The following people were in the queue but didn't have chance to ask questions:
-
-
Li-Hsiang Sun
-
Yunbo Li
-
Kiseon Ryu
-
Jay Yang
-
Rojan Chitrakar
-
Shawn Kim
-
Chunyu Hu
-
Guogang Huang
-
Alfred Asterjadhi
-
Vishnu Ratnam
- Q: Is Beacon-A frame transmitted after group addressed frames or before?
- A: Beacon-A frame is transmitted SIFS after a Beacon frame
- Q: Doesn't AID Bitmap element outside of the Multi-link Traffic Indication element defeat the purpose of using Beacon-A frame for future use to solve the Beacon
bloating issue?
- A: AID Bitmap element is replacement of the TIM element in the current text and the TIM element is outside of the Multi-link Traffic Information element so there
is no difference.
- Q: Can we have both methods? In the Beacon and the Beacon-A frame?
- A: It would be better to have one method instead of having two different methods doing the same function.
- Q: does this new way occupy more air time?
- A: it may if the overhead of AID bitmap element is larger than the overhead of unnecessary Per-link bitmap information of non-AP MLDs or STAs indicated in the
TIM element. But, this proposal solves the beacon bloating problem and legacy STAs are not affected due to Multi-link Traffic Indication element.
- Q: How a STA tell whether there is pending multicast frame other than this frame?
- A: I think it can just follow the existing rule (DTIM Count =0 and Traffic Indicator=1)
- Q: Would this frame makes AP to set the multicast bit in TIM to 1?
- A: I don't think it needs to since the Beacon-A Present Flag subfield in the Cap. Info field=1 will indicate presence of the Beacon-A frame.
Regards,
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
|