Miroslav Lichvar wrote:
On Mon, Apr 18, 2016 at 10:14:20AM +0200, Rob Janssen wrote:
Miroslav Lichvar wrote:
Yes, it seems request to the NIST servers have an extra 60 ms delay,
which makes them appear to run 30 ms ahead of UTC.
In my logs it started approximately in October last year and it wasn't a
sudden switch, but rather a slowly increasing percentage of requests
getting this delay, which got close to 100% by the end of the year.
Maybe they are overloaded?
Could that create a constant delay? If the network or the servers were
overloaded, I'd expect the extra delay to be randomly distributed
across some range. From the (limited) data I have it looks like the
distribution of the delay just shifted by ~63 milliseconds, there is
nothing in between.
Just a case of buffer bloat?
Or it could be that they installed some rate limiting / shaping in the path to
limit the bandwidth
usage and/or the impact of DDOS. When this shaping device has a queue length
equivalent to 63ms of delay, you would see that.
Rob
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool
