Criteria for evaluating upstream server quality - NTP

This is a discussion on Criteria for evaluating upstream server quality - NTP ; I'd like to get opinions as to the weight of specific criteria for choosing upstream servers to get time from. For example, which of these is more important than the others, and why? - round-trip delay; - stratum; - server's ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Criteria for evaluating upstream server quality

  1. Criteria for evaluating upstream server quality

    I'd like to get opinions as to the weight of specific criteria for
    choosing upstream servers to get time from.

    For example, which of these is more important than the others, and why?

    - round-trip delay;
    - stratum;
    - server's own configuration;
    - and so on...

    For now, I've got the following billboard (refid's trimmed):

    mac> ntpq -pn | sort -k8.1,8.6n
    remote refid st t when poll reach delay offset jitter
    ================================================== =======================
    +132.205.7.81 132.205... 3 u 131 1024 377 5.897 -0.615 0.149
    -24.200.225.157 10.23.12.. 4 u 390 1024 377 5.776 -1.854 0.190
    -216.113.122.37 207.253... 3 u 77 1024 377 5.107 1.838 0.721
    +192.26.210.193 172.22.... 2 u 199 1024 377 6.229 -0.214 0.439
    -132.208.250.3 204.123... 2 u 7 1024 377 6.855 1.577 1.506
    *132.246.168.148 132.246... 2 u 70 1024 377 8.820 -0.307 0.764
    -199.212.17.21 142.135... 2 u 32 1024 377 9.314 -7.228 0.411
    -132.214.200.120 18.26.4... 2 u 34 1024 377 17.752 -2.384 1.486
    -209.87.233.53 209.87.... 2 u 205 1024 377 36.763 3.240 0.631

    I concede this may a bit too many servers for domestic use, but how do I
    choose which ones to remove?

    For example, the closest servers to me are actually my ISP's cisco routers
    (24.200.225.157 and 216.113.122.37). Although the round-trip is short
    (about 5-6ms), the routers are polling two time sources, which is a no-no.
    Furthermore, the one at 216 has lost connection to one of its servers for
    close to 4 months, indicating it's probably not monitored. Are those
    trustworthy time sources?

    Some stratum 2 servers I get time from are actually polling stratum 1's
    that have a pretty significant round-trip (more than 70ms), so is this any
    better than having my ntp getting its time from a stratum 3 or 4 that's
    closer to its own upstream sources?

    I already read the standard "best practice" of choosing servers that are
    close-by in the network, but I'd like to learn a bit more about how to
    choose a good-quality time source.

    Opinions are suggestions are welcome...

    --
    Pierre Dubuc
    pldubuc@yahoo.ca

  2. Re: Criteria for evaluating upstream server quality

    Pierre Dubuc wrote:
    > I'd like to get opinions as to the weight of specific criteria for
    > choosing upstream servers to get time from.
    >
    > For example, which of these is more important than the others, and why?
    >
    > - round-trip delay;
    > - stratum;
    > - server's own configuration;
    > - and so on...
    >
    > For now, I've got the following billboard (refid's trimmed):
    >
    > mac> ntpq -pn | sort -k8.1,8.6n
    > remote refid st t when poll reach delay offset jitter
    > ================================================== =======================
    > +132.205.7.81 132.205... 3 u 131 1024 377 5.897 -0.615 0.149
    > -24.200.225.157 10.23.12.. 4 u 390 1024 377 5.776 -1.854 0.190
    > -216.113.122.37 207.253... 3 u 77 1024 377 5.107 1.838 0.721
    > +192.26.210.193 172.22.... 2 u 199 1024 377 6.229 -0.214 0.439
    > -132.208.250.3 204.123... 2 u 7 1024 377 6.855 1.577 1.506
    > *132.246.168.148 132.246... 2 u 70 1024 377 8.820 -0.307 0.764
    > -199.212.17.21 142.135... 2 u 32 1024 377 9.314 -7.228 0.411
    > -132.214.200.120 18.26.4... 2 u 34 1024 377 17.752 -2.384 1.486
    > -209.87.233.53 209.87.... 2 u 205 1024 377 36.763 3.240 0.631
    >
    > I concede this may a bit too many servers for domestic use, but how do I
    > choose which ones to remove?
    >

    NTPD has chosen the best servers. The very best, at the moment the
    above ntpq banner was taken is the one marked with the "*". The runners
    up are marked with "+". Note that this is subject to change, with time.

    I'd drop the last two because of the longer round trip delays. That
    will leave you with seven servers which is still a lot for most
    purposes! You could get along with as few as four.








  3. Re: Criteria for evaluating upstream server quality


    >Some stratum 2 servers I get time from are actually polling stratum 1's
    >that have a pretty significant round-trip (more than 70ms), so is this any
    >better than having my ntp getting its time from a stratum 3 or 4 that's
    >closer to its own upstream sources?


    A long round-trip time isn't evil all by itself. What's bad is
    asymmetric flight times. You get that from either queuing delays
    (busy links) or non-symmetric routing.

    Short round trip times usually have less opportunity for asymmetric
    routing. They can still get long queueing delays.

    --
    These are my opinions, not necessarily my employer's. I hate spam.


+ Reply to Thread