Re: Statistical Sampling of Data- Discussion onrequirement and proposal on advanced featur

Hi,

>You've missed the deadline for submitting -00- drafts.  However,
since
>this does appear relevant to discussion of advancing or updating the
>event and expression MIBs, I'd encourage you to make the URL for
>your work available until the i-d can be posted.
 
 
I will. May be after 4 days because of some problem.


>How would you compare this with what can be done with RFC 2982
>section 2.3.2?  See also the WG archives for May of 2002.  Further
>note that last year we were toying with the idea of making that
document
>"Historic", so I think it's really important for us to understand how
>this approach is better and more likely to be widely implemented.
 
section 2.3.2  functionality as what we are proposing but in our
framework, we do not let wildcard. And hence we monitor either the
scalar object or selective or all instances of a columnaar object.
 
...
> for such parameter.  There is another problem that arises vis-a-vis
the
> history collection of columnar variables. The numbers of instances
for
> these parameters are not fixed. This means that there could be new
> instances that are added or existing ones removed, as per the
dynamic
> behavior of the system being monitored. For example, new interfaces
...

Yes you are correct that the above mentioned phenomena could happend
because of the change in the access control .

>Looking at this from an RFC 2741 perspective, how could it
distinguish
>between, for example, something being deleted and its becoming
>invisible due to access control?  Does this distinction matter,
either
>operationally or mathematically?

While designing we did not consider about the access control but this
is how it should behave,
 
Operationally: No distinction. No matter due to what reason the
instance is not available, best is to stop collecting the history data.
 
Mathematically:  Yes distinct. suppose a instance is deleted it never
gets into the filter criteria where as a invisiable instance is
considered for filtering.
 

>How does this differ from what RFC 2981 provides?
 
I would like to give more thought on this above point before I answer
anything.


>One last general comment:  the phrase "statistical sampling" can mean
>several very different things.  If you think that it is the correct
term for
>what you're proposing, it'd be helpful to define fairly precisely the
>sense in which you intend to use it, particularly due to the
unfortunate
>events associated with the phrase in RMON.

I used "statistical sampling" for history collection. I used
statistical sampling word from RMON, it was my thought that people are
already have associated  history collection to  statistical sampling
andso continue to use. I will give a second thought about the word
"statistical sampling".



Regards,
Shailaja.