I made my thoughts about asymmetric lines some times ago. https://blog.mayer.tv/2016/05/02/ntp-and-adsl-part2.html https://blog.mayer.tv/2016/03/16/ntp-and-adsl.html
@vom513 if you know upload and download speed then mathematics should give you the same result. // Hans — On 03.12.2022, at 20:46, Jan Ceuleers <jan.ceule...@computer.org<mailto:jan.ceule...@computer.org>> wrote: Hans is right, but the huff'n'puff filter can alleviate that, provided that the slower direction of the link is not permanently fully loaded, such that there are times when the latency is the same in both directions. On 3 December 2022 20:22:59 CET, MAYER Hans <hans.ma...@iiasa.ac.at<mailto:hans.ma...@iiasa.ac.at>> wrote: Is your Internet connection symmetric? Is upload and download speed the same? If not this is the reason. // Hans sent from my mobile device ________________________________ From: vom513 <vom...@gmail.com<mailto:vom...@gmail.com>> Sent: Saturday, December 3, 2022 7:53:55 PM To: NTP Questions mailing list <questions@lists.ntp.org<mailto:questions@lists.ntp.org>> Subject: [questions] Local ST 1 (GPS/GNSS) servers are consistently offset from internet ST1/2 Hello all, This has been bugging me for a bit - but not enough till now to post a message :) I have two boxes here at home (this is “hobbyist” grade gear, not big commercial appliances). My hardware is: SyncBOX - https://www.worldtimesolutions.com/products/gps_time_server.html This is connected to a small “Jetway” x86 mini PC via DB9 / real UART serial. UPUTronics Pi Hat - https://thepihut.com/products/raspberry-pi-gps-hat It’s on a Pi obviously (3B I think) Both of these appear to be very stable. I graph the offset in MRTG and they both average about +/- 1-2 us. However - when I compare them each individually to a list of about 10 public ST1 and ST2 servers, they both are averaging about -3.5 ms offset. I did this with ntpdate -q and punched the values into Excel. To also back this data up - I feed a public pool member with these two guys - and my graphs on the server management page also show ~ -3.5 ms offset. (These are the same box…) https://www.ntppool.org/scores/216.143.11.126 https://www.ntppool.org/scores/2607:ff70:11::10 Also - ignore the scatter up above 0. I recently set prefer on both these guys to nail them down and track the pool graphs. So while I guess 3ms isn’t the end of the world, I’m just scratching my head as to why both my boxes are consistent with this offset from the “world”. I suppose I could config both to offset by this amount in their config, but that seems a bit heavy handed. I’d much rather understand why they both converge at this offset. Thanks in advance for any info and clue. PS: Something I just thought of before I hit send… Both of these are behind my home cable modem. And while the jitter is actually pretty good, I know there is inherent latency and jitter in most current DOCSIS networks. I think DOCSIS 3.1 has some really nice bufferbloat / latency / jitter “fixes” (AQM I think it’s called) - but my provider isn’t there yet (and I’m on Docsis 3.0 anyway). Could this be it ?
