Miroslav Lichvar [Thu, 28 Apr 2011 15:32:13 +0000 (17:32 +0200)]
Modify weight calculation again
Dividing the weights by variance or unweighted variance seems to have a
significant negative impact on response with normally distributed
network delays.
Divide by the difference between the mean and minimum distance instead.
It should be stable as there is no loop and the response seems to be a
good compromise between the original minimum distance weighting which
works well with normally distributed delays and the variance weighting
which works well with exponentially distributed delays.
Add configure option to force retry on DNS failure
This is apparently needed on system which keep nameservers specified
in /etc/resolv.conf even when there is no network connection. Should be
used with care as invalid names will be retried forever.
Miroslav Lichvar [Tue, 25 Jan 2011 16:22:00 +0000 (17:22 +0100)]
Make importance of stratum in source selection configurable
Instead of always selecting the source with minimum stratum, add weighted
stratum to the distance when comparing selectable sources. The weight
can be configured with new stratumweight directive and can be set to
zero to ignore stratum completely, by default 1.0.
Miroslav Lichvar [Mon, 24 Jan 2011 17:04:33 +0000 (18:04 +0100)]
Disable variance source test for now
It seems it triggers even with one source alone if it's close and
accurate or polling interval is long enough. It would need to include
max_clock_error.
Miroslav Lichvar [Mon, 24 Jan 2011 16:09:00 +0000 (17:09 +0100)]
Select source with minimum distance using a scoring system
Each source has a score against currently selected source which is
updated (multiplied by ratio of their distances) when one of the two
sources has a new sample. When the score reaches a limit, the source
will be selected. This should allow to slowly select the source with
minimum distance without frequent reselecting.
To avoid switching between sources with very variable distances (e.g. on
LAN or when upstream server uses a longer polling interval), sources
that are currently not selected are penalized by a fixed distance. This
can be configured with new reselectdist directive (100 microseconds by
default).
Miroslav Lichvar [Thu, 16 Dec 2010 17:27:38 +0000 (18:27 +0100)]
Delay selecting source on start
On start, when servers are reachable and use the same polling interval,
wait for them to have the same reachability register (which corresponds
to the number of samples in sourcestats) before selecting one.
Miroslav Lichvar [Thu, 16 Dec 2010 15:55:59 +0000 (16:55 +0100)]
Don't update local reference unnecessarily
Local reference is now updated only when a new source is selected or
match_addr is equal to the selected source, match_addr 0 is no longer
treated specially.
Miroslav Lichvar [Thu, 16 Dec 2010 15:01:29 +0000 (16:01 +0100)]
Track reachability in sources module
Add new flag to source struct to indicate when source is selectable
(packets with good headers are received) and use a reachability
register for last 8 samples instead of the reachable flag. Source
drivers now provide only reachability updates.
Require that the ratio of the increase in delay from the minimum one in
the stats data register to the standard deviation of the offsets in the
register is less than maxdelaydevratio or the difference between
measured offset and predicted offset is larger than the increase in
delay. In the allowed delay increase is included also skew and maximum
clock frequency error.
Miroslav Lichvar [Wed, 13 Oct 2010 14:49:28 +0000 (16:49 +0200)]
Base poll adjustment on number of sourcestats samples
Instead of following skew changes, adjust polling interval so that the
number of measurements used in the regression algorithm remains close to
a target value. It can be configured with a new polltarget option
(6 by default).