[ntp:questions] Proposed NTP solution for a network

Jason bmwjason at bmwlt.com
Fri Mar 6 03:22:09 UTC 2009


Ryan Malayter wrote:
> On Tue, Mar 3, 2009 at 7:21 PM, Jason <bmwjason at bmwlt.com> wrote:
> 
>> I'm pretty sure the appliance might be more capable than the LVC18, same
>> manufacturer as many gov't agencies use, better antenna, highest grade
>> coaxial cable from the antennas, etc. Has PPS and 10Mhz outputs. The
>> problem is getting the PPS into the servers.
> 
> One thing you could experiment with is doing a sort of "poor man's
> PPS" over Ethernet using broadcast NTP mode with the GPS devices as
> servers and as short a poll interval as is possible. If all of the
> blades are running the same hardware, kernel, and NTP versions, using
> broadcast mode would result in NTP clients having essentially the same
> timing information, and so they should behave as similarly as possible
> within the limits of the hardware and environment.
> 
> Ntpd would still have to deal with the non-deterministic load and
> thermal variations on each blade, of course, but such a scheme might
> help with your goal of keeping all the blades as closely synced as
> possible to each other. Obviously to address failure you would need
> three or more GPS devices putting out broadcast packets in each
> broadcast domain, using the prefer keyword on the clients to select
> one as the leader of the pack.
> 
> Anyway, this is just an idea I had in the shower this morning. I am
> sure the actual ntpd developers here will point out many reasons why
> my scheme is not workable. However, it sounds as though you have the
> resources to fully experiment with this - or any other proposed
> solution - in a duplicate hardware environment, correct?
> 

Due to other factors on the LAN, I don't think broadcasting will work 
out, but it is worth looking at -- that would be reminiscent of PTP 
(ieee 1588) and it's multicast (that is, if I've understood the PTP 
protocol correctly -- maybe not).

Yes, we have a lab to experiment in, although we can't generate the same 
system load as in production.

A sys admin and I discussed the possible solutions today, and we have a 
potential winner, although we can't reach the small 10s of uSec goal, we 
can reach several 100s of uSec (and probably less) easily. We are going 
to make a proposal to mgmt to increase the number of S1 servers, the 
number of GPS receivers, and I'm looking for an Rb clock source for the 
main datacenter as well. We are also re-engaging the blade / enclosure 
vendor.

We will also be working on custom monitoring software that will provide 
a very early alert if one of the S0 sources or S1 servers drifts outside 
some bounds. We are also going to start retrieving the various NTP stats 
files to our statistics database(s) so that the software guys can 
incorporate that into the alerting and health monitoring application(s).

Thanks to all for your assistance, I'll post back how things get along.

Jason.




More information about the questions mailing list