Each source is characterized by the offset and delay measured by the on-wire protocol and the dispersion and jitter calculated by the clock filter algorithm of the peer process. Each time an NTP packet is received from a source, the dispersion is initialized by the sum of the precisions of the server and client. The offset, delay and dispersion values are inserted as the youngest stage of an 8-stage shift register, thus discarding the oldest stage. Subsequently, the dispersion in each stage is increased at a fixed rate of 15 ms/s, representing the worst case error due to skew between the server and client clocks. The clock filter algorithm in each peer process selects the stage with the lowest delay, which generally represents the most accurate values, and the associated offset and delay values become the peer variables of the same name. The peer dispersion continues to grow at the same rate as the register dispersion. The peer dispersion is determined as a weighted average of the dispersion samples in the shift register. Finally, the peer jitter is determined as the root-mean-square (RMS) average of all the offset samples in the shift register relative to the selected sample.

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Name Hardware System Precision
deacon.udel.edu Sun Blade 1500 SunOS/5.10 -20
howland.udel.edu i386 2.8 GHZ dual core SunOS/5.11 -21
beauregard.udel.edu i386 2.8 GHZ dual core FreeBSD/8.0 -19
pogo.udel.edu UltraSPARC 5 SunOS/5.10 -20
huey.usel.edu i386 SunOS/5.11 -22
rackety.udel.edu i386/Pentium II FreeBSD/6.1 -18
tick.usno.navy.mil HP ia64 HP-UX/11.31 -20
Meinberg LANtime M600 GPS i586 Linux/2.6.15 -19
time1.chu.nrc.ca i686 Linux/2.6.27 -20

Table 1. Typical Precision for Various Machines

The offset, delay and dispersion values are inserted as the youngest stage of an 8-stage shift register, thus discarding the oldest stage. Subsequently, the dispersion in each stage is increased at a fixed rate of 15 ms/s, representing the worst case error due to skew between the server and client clocks. The clock filter algorithm in each peer process selects the stage with the lowest delay, which generally represents the most accurate values, and the associated offset and delay values become the peer variables of the same name. The peer dispersion continues to grow at the same rate as the register dispersion. The peer dispersion is determined as a weighted average of the dispersion samples in the shift register. Finally, the peer jitter is determined as the root-mean-square (RMS) average of all the offset samples in the shift register relative to the selected sample.

The clock filter algorithm continues to process packets in this way until the source is no longer reachable. In this case the algorithm inserts dummy samples with "infinite" dispersion are inserted in the shift register, thus displacing old samples.

The composition of the survivor population and the system peer selection is redetermined as each update from each server is received. The system variables are copied from the peer variables of the same name and the system stratum set one greater than the system peer stratum. Like peer dispersion, the system dispersion increases at the same rate so, even if all sources have become unreachable, the daemon appears to upstratum clients at ever increasing dispersion.

Name	Hardware	System	Precision
deacon.udel.edu	Sun Blade 1500	SunOS/5.10	-20
howland.udel.edu	i386 2.8 GHZ dual core	SunOS/5.11	-21
beauregard.udel.edu	i386 2.8 GHZ dual core	FreeBSD/8.0	-19
pogo.udel.edu	UltraSPARC 5	SunOS/5.10	-20
huey.usel.edu	i386	SunOS/5.11	-22
rackety.udel.edu	i386/Pentium II	FreeBSD/6.1	-18
tick.usno.navy.mil	HP ia64	HP-UX/11.31	-20
Meinberg LANtime M600 GPS	i586	Linux/2.6.15	-19
time1.chu.nrc.ca	i686	Linux/2.6.27	-20

How Statistics are Determined

Reachability and Selection Criteria

Related Links

New Features

How NTP Works

Table of Contents