Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Hi, John: Sorry for the late reply. Thanks for your question. See inline. Thanks. Chunyu From: Wullert, John R II <jwullert@xxxxxxxxxxxxxxxxx> Chunyu, Thanks for the presentation. I also had some questions:
[Chunyu:] we think majority of applications that requires low latency as well as small jitter can be well modelled as periodic, and bursty; or served so such that they can enjoy a more predictable latency performance. It also has additional power saving benefit For the upstream of AR/VR applications, it also has such characteristics. E.g. IMU, camera, sensor captured data. They may not have exact the same period. But often, the application has to be designed at system level, and a system level of pipeline helps optimize overall performance. The multiple (DL/UL or within either direction) traffic streams can be grouped into one or two or few periodic streams as results. There are quite some literatures online, e.g. [link].
[Chunyu:] agree not all traffic are periodic. The P-slots/schedule are intended for periodic/latency-sensitive traffic that has non-trivial amount of bandwidth requirement. The non-periodic doesn’t need to operate on slot boundary in a time zone where none of slots are P-slots. Efficiency is sacrificed mostly at the R/P-slot boundaries.
[Chunyu:] yes, we can surely consider this and provide means for network operator/AP owners to configure in their deployment. E.g. this can be achieved by AP announces maximum total time allocated to P-slots, and/or maximum # of slots assigned to each TC or STA. Welcome for contributions discussing this aspect. Thanks, John From: Chunyu Hu <chunyuhu07@xxxxxxxxx> Hi, Zhijie: Thanks for the questions. See inline. Chunyu On Wed, Aug 26, 2020 at 7:20 PM Yang, Zhijie (NSB - CN/Shanghai) <zhijie.yang@xxxxxxxxxxxxxxx> wrote:
A: correct. retransmission follows the same rule.
A: For the OBSS, it would have the requirement that the neighboring BSSs understand the same protocol however STAs within OBSS doesn't need to track every slot boundary. Additional OBSS coordination can be designed (see the slot information format design --> information bitmap that contains OBSS) such that BSS informs each other the protected time periods in its own TSF time domain. It's some work we can extend to.
A: in this set of simulation, rate is fixed and every STAs can hear each other. No hidden terminal situation is simulated. We are mostly focusing on the protocol effectiveness here. MSDU agg is ON: 2 MSDUs per A-MSDU. A-MPDU is on and BA window of 64 is used -- for all three schemes (VO uses txop limit 0). We set txop limit to 0 for VO for a fair comparison and mean to demonstrate that using an aggressive CW/AIFSN doesn't solely solve the problem.
A: the requester and granter need to take into account their link operating state: that includes their tx/rx capability, rssi, traffic, to request reasonable amount of time in slots. In practice, one can go through some probing stage to start the initial assignment. It may need to do so with some margin to tolerate traffic/channel dynamics. It may need to adjust assignment as traffic/channel vary over time but hopefully at slow pace.
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 |