It's a short draft so pretty easy to read
", "time": "2025-07-23T14:10:04Z"}, {"author": "Christian Huitema", "text": "The main issue going from PIE to FQ-PIE are time effects. PIE samples every packet. FQ-PIE samples on packets per queue, which can be much less often. How does that play out, exactly?
", "time": "2025-07-23T14:11:31Z"}, {"author": "Christian Huitema", "text": "Previous IETF show that this was an issue with FQ-CODEL -- it took a long time to ramp up the estimate of the dequeue rate, and thus data rate was only growing very slowly after bandwidth increased. DO we have the same issue with FQ-PIE?
", "time": "2025-07-23T14:14:02Z"}, {"author": "Mohit Tahiliani", "text": "Hi Christian, the data rate with FQ-PIE seems to grow fine when the bandwidth is increased (Slide 14 in the PDF of the talk), atleast in the experiments that we performed during the hackathon in IETF 122
", "time": "2025-07-23T14:18:19Z"}, {"author": "Christian Huitema", "text": "The other potential issue with PIE is the sensitivity to delay jitter, e.g., the kind of jitter that we see in Wi-Fi networks when several adjacent networks use the same channels. This happens in apartments building, with multiple tenants all running their own Wi-Fi networks, and also in office buildings hosting multiple businesses. Standard latency would be 4ms, but jumps to 50ms or even 200ms can happen, due to the exponential backoff algorithms used for L2 packet retransmission. I wonder whether we have simulations of that with PIE. I suspect this would affect estimates based on time stamps.
", "time": "2025-07-23T14:28:38Z"}, {"author": "Anna Brunstrom", "text": "In the previous session Greg mentioned that there was a paper on the TCP fallback in ANRW yeasterday. Here is a link to the paper: https://dl.acm.org/doi/pdf/10.1145/3744200.3744770 (should you have problem with access you can reach it from the ANRW program page https://www.irtf.org/anrw/2025/program.html)
\nThe recordings from ANRW is also available in the IETF agenda with multiple presentations on Transport.
^TCP Prague fallback
", "time": "2025-07-23T14:33:01Z"}, {"author": "Mohit Tahiliani", "text": "@Christian - No, we haven't tested PIE in such a scenario. I'll make a note of this scenario and give it a try while performing the next set of experiments. Thanks!
", "time": "2025-07-23T14:38:57Z"}]