[ntp:questions] Surplus Alpha as timeserver?

Tom Smith smith at cag.lkg.hp.com
Sat Mar 26 03:06:27 UTC 2005


Brad Knowles wrote:
> At 10:23 PM +0000 2005-03-25, Bjorn Gabrielsson wrote:
> 
>>  Whats the general opinion. Should a public server be running nano
>>  kernal? or is that an overkill?
> 
> 
>     I'm curious.  I am reasonably familiar with the kernel architecture 
> of *BSD, and I'm pretty sure I have the basic knowledge of how the Linux 
> kernel architecture is laid out.  None of these involve micro or nano 
> kernels.
> 
>     What are you thinking of, and how do you think that this will help 
> the security situation?
> 
> 
>     Note that OpenBSD now ships with their own NTP server OpenNTPd. 
> While I've said some pretty nasty things about it in the past, and while 
> I recognize that Darren has been working on making improvements, I do 
> not believe that OpenNTPd is yet ready for production use on a public 
> server.
> 
>     I'm not sure how much luck you're going to get with the Reference 
> Implementation on OpenBSD, and I know you're not going to get any 
> support from the OpenBSD guys if you should happen to have any problems.
> 

I'm really not sure what the kernel implmentations do or do not do
w.r.t security except that they move a lot of the details of PLL
maintenance out of user space, but of the possibilities you are
considering, Tru64 implments a pre-nano kernel. "options NTP_TIME"
together with "options MICRO_TIME" in the kernel configuration file
enables, respectively, the kernel PLL and microsecond interpolation
between the kernel ticks of 1/1024 seconds. The former option
is listed as an option when you build the kernel. The latter
you have to edit in by hand. Together with NTP V4 and a good
set of upstream servers or a reference clock, they
generally keep the clock strapped down to within a few hundred
microseconds or better of what NTP thinks the time ought to be.

-Tom



More information about the questions mailing list