Hi Brennan, really cool stuff! On May 13, 2022, at 01:35, Brennen Smith > wrote: Hey all, Here's some details about the final figures: 1. Loaded latency is simply an extension of the latency stage - same methodology and medium Where could I find a technical description of that method? 2. We use a "warmed up" TCP connection for latency Is that one of the load bearing flows or a non load bearing parallel flow that has successfully finished the TCP handshake? 3. We use the interquartile mean for the "displayed value" I wonder whether reporting something like a 95 or 99% quantiles (so either 1& and 99%, or 5% and 95%) in addition to min and max and interquartile mean would be possible? And since you use the 25 and 75% values already, maybe these as well? Awesome seeing those numbers in the wild and looking forward to bringing more awareness to this issue. For sure, so far many discussions about debugging/reducing bufferbloat started with people wanting to use the ubiquitious speedtest.nat nodes (where finding a close by one typically is easy/possible); and now this will actually work (assuming they use Android or iOS clients). Question: will the apps for the other OSs as well as the CLI and the web version also be changed to include these numbers in the foreseeable future? Kind Regards Sebastian Cheers, Brennen Smith VP Technology (206) 739-0807 | brennen@ookla.com linkedin.com/in/brennensmith On Thu, May 12, 2022 at 4:18 PM Dave Taht wrote: Yea. peak 2500ms latency on download, 600ms on upload, that's about right. :/ The upload figure appears to be a bit lower (better) than what I measured 1 year ago, here (for the 30 or so new subscribers on this list, this was why I started it): https://docs.google.com/document/d/1puRjUVxJ6cCv-rgQ_zn-jWZU9ae0jZbFATLf4PQKblM/edit I don't know if ookla are throwing out arp related stuff, or using the 99th percentile to calculate the final figure. I hope to learn more about their calculations in the coming weeks. I do keep hoping we can get more updated flent data over longer intervals than 20s. With starlink changing their allocation scheme every 15s, what number do you pick? And like I said, some packet caps of this app would be good, too. On Thu, May 12, 2022 at 4:07 PM Nathan Owens wrote: Here's a test I did today: https://www.speedtest.net/my-result/i/5112435305 [cid:180ba86ed9c36eb9f291] On Thu, May 12, 2022 at 4:04 PM Dave Taht wrote: Ookla has just put out an ios and android app that also continuously measures latency under load. I'm interested in calibrating the results of that vs a vs flent benchmarks verses starlink. Can someone here give this a shot? More details here: https://www.ookla.com/articles/introducing-loaded-latency IDEALLY, this would be over wifi -> starlink , with a packet capture in the middle. But I'd settle for a few test results from starlink folk of this new app, first. -- FQ World Domination pending: https://blog.cerowrt.org/post/state_of_fq_codel/ Dave Täht CEO, TekLibre, LLC _______________________________________________ Starlink mailing list Starlink@lists.bufferbloat.net https://lists.bufferbloat.net/listinfo/starlink -- FQ World Domination pending: https://blog.cerowrt.org/post/state_of_fq_codel/ Dave Täht CEO, TekLibre, LLC This email, its contents and attachments contain information from Ziff Davis, Inc. and/or its affiliates which may be privileged, confidential or otherwise protected from disclosure. The information is intended to be for the addressee(s) only. If you are not an addressee, any disclosure, copy, distribution or use of the contents of this message is prohibited. If you have received this email in error, please notify the sender by reply email and delete the original message and any copies. _______________________________________________ Starlink mailing list Starlink@lists.bufferbloat.net https://lists.bufferbloat.net/listinfo/starlink