--===============0074867417==
Content-Type: multipart/alternative;
boundary="----=_Part_77563_26618184.1177678355632"

------=_Part_77563_26618184.1177678355632
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

On 4/24/07, Dave Shield wrote:
>
> On 24/04/07, Nguyen Huy Ha wrote:
> > > Hi, Is it possible to get timestamps of responses right when they

> leave
> > > their agent, and how?

>
> You could turn on packet dumps, which by default would get logged in
> the agent's syslog - complete with timestamp.
>
>
> > It should be impossible since there is no timestamp field in SNMP/UDP/IP
> > stack.

>
> It's certainly true that timestamps are not part of the SNMP protocol.
> This is deliberate - since SNMP may well be implemented on fairly
> low-level hardware, where they may not be any form of externally
> reliable timestamp.
> That's why the fundamental time measurement for most MIBs is
> 'sysUpTime' - the length of time that the agent has been up and running.
> This is purely self-contained, and doesn't rely on any external clock.
>
> > I was just wondering that the counter values that we get from agents
> > would be more precise if we have also the timestamp at the agents since

> it
> > also takes a while for responses from agents to get to their manager.

>
> Quite honestly, I suspect that delays in retrieving the counter values
> from the underlying subsystems would be of equal significance to
> network delays. Even more so when it comes to any internal caching
> of data within the agent.
> All that a timestamp would give you would be the time that a response
> was sent - which is not necessarily the time that the individual counters
> were actually sampled.
>
> Dave
>


Yes, it'd probably be equal significance between network delays and agent's
processing delays. But we will get more precise result if we can reduce as
many variables, especially in large networks. Cisco has IP-SLA feature that
takes timestamp whenever interesting packets arrive or leave a router's
interface. I think that is a nice feature.

Br.
Ha

------=_Part_77563_26618184.1177678355632
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline


On 4/24/07, Dave Shield <.T.Shield@csc.liv.ac.uk">D.T.Shield@csc.liv.ac.uk> wrote:

On 24/04/07, Nguyen Huy Ha <ha.h.ngu@gmail.com> wrote:
> > Hi, Is it possible to get timestamps of responses right when they leave
> > their agent, and how?


You could turn on packet dumps, which by default would get logged in
the agent's syslog - complete with timestamp.


> It should be impossible since there is no timestamp field in SNMP/UDP/IP
> stack.


It's certainly true that timestamps are not part of the SNMP protocol.
This is deliberate - since SNMP may well be implemented on fairly
low-level hardware, where they may not be any form of externally

reliable timestamp.
   That's why the fundamental time measurement for most MIBs is
'sysUpTime' - the length of time that the agent has been up and running.
This is purely self-contained, and doesn't rely on any external clock.


>   I was just wondering that the counter values that we get from agents
> would be more precise if we have also the timestamp at the agents since it
> also takes a while for responses from agents to get to their manager.


Quite honestly, I suspect that delays in retrieving the counter values
from the underlying subsystems would be of equal significance to
network delays.   Even more so when it comes to any internal caching

of data within the agent.
   All that a timestamp would give you would be the time that a response
was sent - which is not necessarily the time that the individual counters
were actually sampled.

Dave


Yes, it'd probably be equal significance between network delays and
agent's processing delays. But we will get more precise result if we
can reduce as many variables, especially in large networks. Cisco has
IP-SLA feature that takes timestamp whenever interesting packets arrive
or leave a router's interface. I think that is a nice feature.



Br.

Ha


------=_Part_77563_26618184.1177678355632--


--===============0074867417==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
--===============0074867417==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/...et-snmp-coders

--===============0074867417==--