Hello David, >>> I'm surprised you don't trust "pool" servers. My experience is that using >>> the pool directive, and allowing NTP to expand its server list automatically >>> to the maximum number of servers, gives good results usually with one of two >>> servers at least being stratum-1 GPS-locked.
The main reason I do not trust "pool" servers is because there is no guarantee of which server you will get. I might be paranoid, but I am worried about rogue servers, and I much rather trust well known public stratum-1 NTP servers. So mainly a matter of trust. As far as their quality, I've actually never really seen public stratum-1 servers being overloaded. A good number of them are actually a group of servers and clients round-robin among all of them. For instance if I lookup the DNS records of the stratum-1 NTP severe closer to me (University of Washington, Seattle), I see it is actually 3 servers: fcattane@linux-mint-64:~$ dig bigben.cac.washington.edu ;; ANSWER SECTION: bigben.cac.washington.edu. 5 IN CNAME time.u.washington.edu. time.u.washington.edu. 4 IN A 140.142.234.133 time.u.washington.edu. 4 IN A 140.142.1.8 time.u.washington.edu. 4 IN A 140.142.2.8 >>> If you have a PPS feed from that GPS board, you could easily add it to a >>> Linux PC (Raspberry Pi, for example) or even a Windows box and use that >>> locally as your own stratum-1 server. Yes, all three stratum-1 servers I run at home all have a PPS output from GPS (either vanilla GPS receiver or GPSDO) which I feed to NTPD via serial port DCD input. >>> I never had much success with >>> peering - it seemed that when one server had a higher offset for whatever >>> reason it dragged the other with it. I use multiple independent stratum-1 >>> servers - one a Linux X86, one a Raspberry Pi, and one LeoNTP box. That's interesting you had experience with peering. My own experience with it was pretty good, but in my case I did the work to set up internal time sync for the cloud computing megacorp I worked for, and thus I had the luxury of choosing among several hundred servers and I picked the ones which had the lowest PPM drift. I wonder if NTPD's peering algorithm runs into trouble if the clock jitter between machines is too high -- hard to say without doing more experimenting. My experience with peering was good but only lasted about 6 months, as few months later we actually bought some Symmetricom stratum-1 boxes to avoid external dependencies and certify NIST traceability. The obvious advantage of peering is that if you don't have your own stratum-1 sources you can continue to serve time even if the internet connection goes down. Of course these days it's so incredibly cheap -- not to mention fun -- to build your own stratum-1 server with cheap GPS receiver with PPS output using either a Rasperry PI, a cheap old X86 PC, or (as I do for 2 of my servers) a pcengines box. Luckily both of us went down this route for our own NTPD servers :-) Cheers ! -- Fio Cattaneo Universal AC, can Entropy be reversed? -- "THERE IS AS YET INSUFFICIENT DATA FOR A MEANINGFUL ANSWER." On Sat, Oct 19, 2019 at 9:04 AM David J Taylor via time-nuts <time-nuts@lists.febo.com> wrote: > > Hi David, I don't particularly trust NTP servers from pool.ntp.org (I > assume that is what you mean by "pool"), and I use public stratum-1 > servers chosen from a public list. Of course I make sure that my usage > complies with the policies and terms of use for each server (some > allow regional use only, some say do avoid using iburst keyword, some > require prior permission and/or notification). > I have found that using 3 to 5 public stratum-1 servers works very > well, and gives a time synchronization which is within 3 to 5 > milliseconds when compared with a reference timing board with GPSDO. > This time offset discrepancy is actually due to the fact that my ISP > (comcast cable) has asymmetric send/receive delays. It disappears when > I tried this setup at my office, which has symmetrical fiber optic > connection to internet. > > Your point about avoiding having all the internal machine hit > stratum-1 servers is a good one however. To avoid that in my setup I > designate 3 machines which serve as an internal stratum-2 pool for > internal distribution. Each of them has 3 to 5 external NTP stratum-1 > servers and they all peer with each other. Then every other internal > machine uses these 3 machines. I have more details on my setup in my > reply to Eric. > > Kind regards, > > -- Fio Cattaneo > ===================================== > > Fio, > > I'm surprised you don't trust "pool" servers. My experience is that using > the pool directive, and allowing NTP to expand its server list automatically > to the maximum number of servers, gives good results usually with one of two > servers at least being stratum-1 GPS-locked. That's using UK and NL > servers. I suppose if you have a poor or overloaded internet connection > server quality doesn't matter as much - well, almost. My ISP is 200/20, and > used to be 200/12. Talk about asymmetric! > > If you have a PPS feed from that GPS board, you could easily add it to a > Linux PC (Raspberry Pi, for example) or even a Windows box and use that > locally as your own stratum-1 server. I never had much success with > peering - it seemed that when one server had a higher offset for whatever > reason it dragged the other with it. I use multiple independent stratum-1 > servers - one a Linux X86, one a Raspberry Pi, and one LeoNTP box. > > Cheers, > David > -- > SatSignal Software - Quality software for you > Web: http://www.satsignal.eu > Email: david-tay...@blueyonder.co.uk > Twitter: @gm8arv > > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there. _______________________________________________ time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.