[time-nuts] Timing performance of servers

[Magnus Danielson]

Fellow time-nuts,

When spending time on a conference last week, I heard one interesting 
comment that they lost data due to bad timing on their Windows servers.

Now, I know that the standard Windows uses SNTP in order to achieve the 
goal of having the timing of the machines sufficiently aligned to allow 
Kerberos authentication. SNTP suffice for that, as it needs to be a 
handful of minutes in line.

If you need better performance than that, you should use NTP (and then 
download and install Meinbergs Windows-client for NTP).

Then again, I would point out that for this type of data, it would most 
probably be better served on a Linux box.

What should be a nice wake-up call for them would be a summation of how 
different strategies would give them clock precision of sufficient 
grade. So, does anyone know of such measurements presented anywhere?

There are bits and pieces, but the ideal for this case would be if they 
where collected in one page/paper.

This is an awareness thing, so that people can do a little more 
well-informed choices.

Cheers,
Magnus