Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Hi Ming,
As stated earlier, SP#198 only proposes a high level principle of utilizing non-primary channels in 802.11 when the primary channel is busy. This very premise ensures fairness to legacy 802.11 devices while improving 802.11 channel utilization (by removing the dependence on the availability of a single primary channel).
Regarding the next level of details on which you have concerns, we have time to discuss and jointly design a scheme that is efficient and agreeable to all. We have copied below your questions/observations and embedded our responses to address them.
Ming>> I also did homework with our colleagues. Based on ETSI regulation, there are two options such that NRU has two types for multi-channel access, one is type A and the other is type B. but the device can just choose one. Response>> You may be referring to the following text in the LAA/NRU specification (37213-g20 section 4.1.6) “An eNB/gNB can access multiple channels on which transmission(s) are performed, according to one of the Type A or Type B procedures”. Here, the phrase “one of” does not mean that a device can select “only one” Type A or Type B. It rather means that a device can select “any one” of the options for any particular channel access attempt and for any particular set of channels. It can even run both the options in parallel and access the channels using whichever option wins. The only constraints are in Type B where a) the channels must be a part of the bonded set and b) the full EDCA channel cannot change more than once a second or must be chosen uniformly randomly. The phrase “one of” has been used in many other places of the LAA/NRU specification with a similar implication. We can share those examples with you if you are interested. Please let us know. We have been actively involved in the formulation of channel access mechanisms in LAA/NRU as well as ETSI regulations and hence, know this. Ming>> Our NRU colleagues told me both type A and type B have their own advantages and drawbacks, it is difficult to say which one is better. Response>>We too aren’t claiming that LAA/NRU Type A (or ETSI Option 1) is always better than LAA/NRU Type B (or ETSI Option 2, which is also the current 802.11 channel access scheme). Rather, we are proposing in 0363r3 to have ETSI Option 1 as an additional channel access option available to 802.11, such that it can be used in situations where it provides an advantage. For example, in our contribution, we have proposed the following scheme: · Step 1: Attempt ETSI Option 2 on a set of 20 MHz channels. This is the current 802.11 channel access scheme. · Step 2: If the channel on which full EDCA is performed finds the channel busy i.e. ETSI Option 2 fails, then perform the following Steps. · Step 3: Attempt Type A on the remaining channels of the set to check if any other combination of channels is available. · Step 4: If any such subset of channels is available, then transmit on them and only for the duration the full EDCA channel in Step 2 is estimated to be busy. · Step 5: At the end of such transmission and/or at the end of the estimated duration up to which the full EDCA channel in Step 2 is estimated to be free, return to the full EDCA channel to start EDCA again. Observation 1: The scheme proposes to look for other free channels to transmit only when the current 802.11 channel access scheme finds the primary (i.e. the full EDCA channel) busy, and only for the duration the primary channel is estimated to be busy. So, while the current 802.11 scheme does not transmit whenever the primary channel is found to be busy (irrespective of the busy/free status of the other channels), the proposed scheme will utilize the estimated busy duration of the primary channel to look for other free channels and transmit on them, if any are found to be free. So, the proposed scheme will always provide higher channel access opportunities for 11be and so will always lead to better performance.
Observation 2: The scheme does not propose to switch the 802.11 primary channel. Rather the scheme proposes that a device look for transmission opportunities on other channels when the primary channel is found to be busy and return to accessing the primary channel before or at the end of the estimated busy duration.
The merits of ETSI Option 1 and Option 2 were discussed during the standardization of LAA in 3GPP in Aug/Oct 2015. NRU carried over this LAA baseline without change. You can refer to the contribution R1-155655 to RAN1 which discusses this in detail. In this submission Alt 1 is Option 2 and Alt 2 is Option 1. Cat-4 LBT is LAA terminology for full EDCA i.e. EDCA with truncated exponential backoff. You can also refer to contributions R1-155458 and R1-155898 for similar discussion and conclusions.
Ming>> As u said, if going to opt 1, then you can not have primary channel and can not use PIFS access (25 us), you lose opportunities to construct the large bandwidth. However, based on your contribution, there is a primary channel. If going to opt 2, the primary channel can not be changed in 1 second. Please tell me how you get the gain under this condition- perform backoff on the secondary channel. Response>> As described in Step 1 - Step 5 above, the device looks for channel access opportunities on other sets of channels only when it determines the primary channel to be busy and only for the duration the primary channel is estimated to be busy. In the current 802.11 scheme, if the primary channel is found to be busy, the device would not have transmitted. In exactly this situation, the proposed scheme provides the device an opportunity to look for other sets of channels and transmit on them if any are found to be free (for example using ETSI Option 1). So, the proposed scheme always provides a higher channel access opportunity compared to the current 802.11 channel access scheme and so always provides a gain. Importantly, the scheme does not propose to switch the primary channel. It only looks for other channel sets once the primary channel is determined to be busy and returns back to accessing the primary channel once the estimated busy duration is over. Regards, Sindhu
From: Ganming (Ming) [mailto:ming.gan@xxxxxxxxxx]
Hello Sindhu,
I also did homework with our colleagues. Based on ETSI regulation, there are two options such that NRU has two types for multi-channel access, one is type A and the other is type B. but the device can just choose one. Our NRU colleagues told me both type A and type B have their own advantages and drawbacks, it is difficult to say which one is better. As u said, if going to opt 1, then you can not have primary channel and can not use PIFS access (25 us), you lose opportunities to construct the large bandwidth. However, based on your contribution, there is a primary channel. If going to opt 2, the primary channel can not be changed in 1 second. Please tell me how you get the gain under this condition- perform backoff on the secondary channel.
Best wishes, Ming Gan
发件人: Sindhu Verma [mailto:000011381223f2e2-dmarc-request@xxxxxxxxxxxxxxxxx]
Hi Ming and Guaogang,
To add something to what Matt/Vinko/Kaiying already mentioned:
Many of the above points have been already mentioned in our contribution 20/0363r3, during the meeting Q&A and in email discussions. It would have also helped to receive this request for discussion a bit earlier. We could then have discussed and resolved these and reached a conclusion before the motion today.
Regards, Sindhu
From: Matthew Fischer [mailto:00000959766b2ff5-dmarc-request@xxxxxxxxxxxxxxxxx]
Ganming,
Please see embedded comments:
On Wed, Sep 16, 2020 at 7:25 PM Ganming (Ming) <ming.gan@xxxxxxxxxx> wrote:
All of the proposals discussed thusfar have indicated that the secondary channel access should be limited to the duration of the PPDU or TXOP on the primary channel. If the busy is caused by a non-802.11 signal of which the duration is unknown, then the secondary access would not occur. If any 802.11 signal has caused that busy condition, then there would be no other 802.11 signal to receive while the first signal is present. [There is no such thing as preemption in 802.11]
Again, none of the proposals that have been discussed would provide immediate access to the secondary channel. Channel state assessment and would include randomization, as without it, collisions will be generated among the competitors for the unused resource.
These existing techniques all rely on primary channel access. If the primary channel is busy, these techniques do nothing to utilize available bandwidth.
Competing technologies in the unlicensed bands have protocol definitions and implementations that allow parallel multi-channel access to be able to find and utilize any and all available channels. If we do not adopt similar techniques, we will not compete well for the use of the resources and our technology will be replaced by something that is better able to take advantage of the open public resource.
This is an R2 feature, allowing plenty of time to ensure that the details are well-vetted.
Matthew Fischer
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 |
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature