[ntpwg] KISS codes

[Greg Dowd]

A subject near and dear to my heart.  The server can have a
configuration for load limit and start sending kiss codes with either a
random or specified list.  Internal hash trees can sort by load in this
"panic" mode.  Either way, we could migrate away from the client based
mgmt of a server resource.  I'm interested in Dave's feedback on the
consequences of forcing larger populations into an undersampled state,
although it is debatable whether a slammed server is providing adequate
service.  This should, however, allow the server to provide the best
service possible under varying loads.  Properly configured clients
should transition to alternate resources.

Another interest of mine is the ability to detect whether a request
frame, at least in some mode, is coming from a ntp termination point.
An example is a sntp client which loads 0.0 into tx.  A possibility
would be something like a high stratum number or an unsynchronized li
field.  This would allow the server to better scale ntp resources.  


Greg Dowd
gdowd at symmetricom dot com (antispam format)
Symmetricom, Inc. 
www.symmetricom.com
"The current implementation is non-obvious and may need to be improved."




-----Original Message-----
From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.org
[mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.org] On Behalf Of
Brad Knowles
Sent: Monday, July 09, 2007 9:51 AM
To: Brad Knowles; anthony.flavin at bt.com; mayer at gis.net;
karen.odonoghue at navy.mil
Cc: ntpwg at lists.ntp.isc.org; stenn at ntp1.ntp.isc.org
Subject: Re: [ntpwg] KISS codes

On 7/9/07, Brad Knowles wrote:

>  So, you cut the problem into chunks, and you solve them as best and  
> as quickly as you can.

I'm assuming that the server has some intelligence so that when it gets
enough load that it is sending out a lot of RATEs, it can go into a
"panic" mode and start sending out KoDs to everyone?


On that same line of reasoning, do we want to enforce a mandatory
exponential backoff for all clients during the initial volley, if they
don't get responses to their queries?  This would help the clients
automatically reduce the load that they present to sites in a UWisc-type
case, and hopefully reduce the need for the victim site to be sending
out KoDs to everyone.

Of course, this would have to be built into both the NTP and SNTP
standards, otherwise most clients that would be likely to be
misconfigured to cause these sorts of situations would not incorporate
this feature.

--
Brad Knowles <brad at shub-internet.org>, Consultant & Author LinkedIn
Profile: <http://tinyurl.com/y8kpxu> Slides from Invited Talks:
<http://tinyurl.com/tj6q4>

09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0
_______________________________________________
ntpwg mailing list
ntpwg at lists.ntp.org
https://lists.ntp.org/mailman/listinfo/ntpwg