lulu wrote:
> Actually I want to know how frequently a system's time get get updated
> by any NTP server.
> I didn't found in man pages that it can be done with user
> interference. So I think this is something done by deamon itself. And
> the frequency is hardcoded inside its source code.
>
> If it is so it may happen that it will over load the cpu. Because it
> must be updating itself with some
> transmission and recepton of packets from server.
>
> So how to know that how frequently a system get itself updated with
> that of a NTP server and is it uniform ?
>
> Is there any chance of getting a cpu overloaded due to this updation
> or due to any operation of xntpd deamon ?
>
> thanks in advance.

[simple explanation]

An NTP client (by default) will ask each server for a new time every 64
seconds when it starts up. As the client time becomes correct, the
interval between requests will gradually lengthen until it reaches 1024
seconds. It is not hard-coded, but dynamic and adaptive, and will reduce
again if needed.

There is no chance of CPU overload.

David

Is this decision of changing the interval dynamically done by any
algorithm
present in rfc related to NTP.

Can't the user interfere in changing the time interval in which a
NTP-client sends or recieve packets

And how of sure u are saying it will never overload the CPU. Because
sending a packet every minute can make a system overloaded at times
also if it is side by side working on
any resource hunting process. So what i mean to askcan't a user control
th packaet sending and recieving interval manually.

lulu wrote:
> Is this decision of changing the interval dynamically done by any
> algorithm
> present in rfc related to NTP.

I believe the algorithm is documented in Dr Mills papers.

http://www.eecis.udel.edu/%7emills/ntp.html

> Can't the user interfere in changing the time interval in which a
> NTP-client sends or recieve packets

You can set the minimum and maximum intervals.

> And how of sure u are saying it will never overload the CPU. Because
> sending a packet every minute can make a system overloaded at times
> also if it is side by side working on
> any resource hunting process. So what i mean to askcan't a user
> control th packaet sending and recieving interval manually.

In my own experience I have never seen anything remotely resembling an
overload on a client system. If a server has thousands of clients, the
load on the server may need to be considered.

David

lulu wrote:

> Is this decision of changing the interval dynamically done by any
> algorithm
> present in rfc related to NTP.
>
Yes. The math is way over my head but it's there

> Can't the user interfere in changing the time interval in which a
> NTP-client sends or recieve packets
>
Yes, but it's a bad idea if you want good timekeeping. The optimum
polling interval changes with network behavior and the behavior of your
local clock.

> And how of sure u are saying it will never overload the CPU. Because
> sending a packet every minute can make a system overloaded at times
> also if it is side by side working on
> any resource hunting process. So what i mean to askcan't a user control
> th packaet sending and recieving interval manually.

Others may be able to offer something like proof but I would guess that
ntpd would use less than one percent of just about any CPU manufactured
in the last five to eight years.

>

Thanks a lot for the answers

I have one more doubt. Let's say I have two cards in one system out of
which one connected to internet and synching its time appropriately
with stratum-2 server.
But I have connected another two system through another card. So right
now I make
theses two machine behave the machine having two cards as a NTP server
for them. ( By changing the ntp.conf as SERVER < INTERNET CONNECTED
IP> )

Made some changes to routing table so that these to local machines who
are connected
2nd card( not connected to internet) can ping to the ist interface.

But still these two servers are not sysncing teir time .

"lulu" wrote in message
news:1164801470.404384.20580@j72g2000cwa.googlegro ups.com...
> Is this decision of changing the interval dynamically done by any
> algorithm
> present in rfc related to NTP.
>
> Can't the user interfere in changing the time interval in which a
> NTP-client sends or recieve packets
>
> And how of sure u are saying it will never overload the CPU. Because
> sending a packet every minute can make a system overloaded at times
> also if it is side by side working on
> any resource hunting process. So what i mean to askcan't a user control
> th packaet sending and recieving interval manually.
>

lulu, if it's any help, my Stratum 2 pool.ntp.org server is a Linux system
running on an Athlon XP 2400+ processor. It suffers no noticable load under
normal conditions, which is about 1500 clients, but even when the server
shows up in the NTP pool DNS, pushing client levels very high (I've seen it
as high as 9731 clients), the load was minimal, all according to Wayne
Schlitt's scripts. Network traffic is really not that much either, ~5 or 6
KB/sec under very high client levels if I remember correctly.

Dennis

On 2006-11-29, lulu wrote:

> Let's say I have two cards in one system out of which one connected to
> internet and synching its time appropriately with stratum-2 server.
>
> But I have connected another two system through another card.
>
> So right now I make theses two machine behave the machine having two
> cards as a NTP server for them. ( By changing the ntp.conf

The ntp.conf on the clients.

> as SERVER < INTERNET CONNECTED IP> )
>
> Made some changes to routing table

ntpd binds all of the interfaces. So you can point your clients at the
IP address of the interface they are connected to. Then you won't have
to complicate your configuration with unnecessary routing.

> But still these two servers are not sysncing teir time .

Is your server synced to a source of time (e.g. remote time server or
local ref-clock)?

Is port 123/UDP open on both your server and its clients?

Are there any restrictions which may be blocking ntp packets between
your server and its clients?

Have you tried pointing your clients at the server's internal interface?

--
Steve Kostecke
NTP Public Services Project - http://ntp.isc.org/

Steve Kostecke wrote:
> On 2006-11-29, lulu wrote:
>
> > Let's say I have two cards in one system out of which one connected to
> > internet and synching its time appropriately with stratum-2 server.
> >
> > But I have connected another two system through another card.
> >
> > So right now I make theses two machine behave the machine having two
> > cards as a NTP server for them. ( By changing the ntp.conf
>
> The ntp.conf on the clients.
>
> > as SERVER < INTERNET CONNECTED IP> )
> >
> > Made some changes to routing table
>
> ntpd binds all of the interfaces. So you can point your clients at the
> IP address of the interface they are connected to. Then you won't have
> to complicate your configuration with unnecessary routing.
>
> > But still these two servers are not sysncing teir time .
>
> Is your server synced to a source of time (e.g. remote time server or
> local ref-clock)?
>
Sorry in the line "two server are not synching ..." infact it is the
two clients are not synching their time. And my server system is
sysnching to a startum-2 server in the net as said before. That
interface

> Is port 123/UDP open on both your server and its clients?
>
I couldn't able to get port is open means what ? As I guess u mean to
ask whether both cleint and server using that port or not , then it is
yes .
> Are there any restrictions which may be blocking ntp packets between
> your server and its clients?
There is no restriction ..

> Have you tried pointing your clients at the server's internal interface?
>
Inernal interface , i think u mean to say the local interface which is
not connected to internet for the stratum-2 server .
> --
> Steve Kostecke
> NTP Public Services Project - http://ntp.isc.org/