"The NTP network is a forest of hosts(...)" - NTP

This is a discussion on "The NTP network is a forest of hosts(...)" - NTP ; Hi, a "forest" being a "set of trees", I have a question with the statement cited: I doubt that NTP sync graph is tree-like (each node having eactly one parent, and the graph having no cycles). Especially with "peer" synchronization ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: "The NTP network is a forest of hosts(...)"

  1. "The NTP network is a forest of hosts(...)"

    Hi,

    a "forest" being a "set of trees", I have a question with the statement cited:
    I doubt that NTP sync graph is tree-like (each node having eactly one parent,
    and the graph having no cycles). Especially with "peer" synchronization with
    peers being at the same startum.

    The statement was taken from page 7 of "NTP Security Model", dealing with the
    basics of autokey. Assuming a tree graph makes the understanding more simple,
    but reality makes implementation quite fuzzy: Which nodes on a group do need
    "CA certificates" (trusted), especially if the topmost level of the group are
    multiple peers at the same stratum.

    As I understood it, NTP propagates trust "down the stratum" (low numbers to
    higher numbers). When initially staring the servers, it's unsure which one
    will sync first, and thus have the lowest startum (and therefore the need for
    the trusted certificate).

    Did I mis-interpret anything?

    At the moment one statum-2 peer has a flash of 408 (no_access, not_proventic,
    ....) while the other peer (with a disconnected refclock) has flash of 608
    (no_access, bad_autokey, not_proventic)

    Regards,
    Ulrich

  2. Re: "The NTP network is a forest of hosts(...)"

    "Ulrich Windl" wrote in message
    news:86lkrhy4l6.fsf@pc8032.rz.uni-regensburg.de...

    > a "forest" being a "set of trees", I have a question with the statement
    > cited: I doubt that NTP sync graph is tree-like (each node having eactly
    > one parent, and the graph having no cycles). Especially with "peer"
    > synchronization with peers being at the same startum.


    It might mean that the _sync_ graph is a tree. At every moment in time,
    only one association can be the selected one.

    Groetjes,
    Maarten Wiltink



+ Reply to Thread