Conference azur::mcc

We are evaluating the development of some DECmcc modules for a customer
project. One of the modules is a new "Performance Analyzer" that will support
their new entities. We are also developing AMs for these new entities.

One of the entities is conceptually similar to the CIRCUIT (supported by the
CIRCUIT AM), that is, it is a "virtual entity" that groups "logically" other
entities (comm. lines).

The question is: Can a "new" PA FM support statistics calculation on those
entities ? in other words, could a new version of the current PA FM support
for example the CIRCUIT entities?

>>>> What is the "new" version of PA FM? Will it have formulae to support 
>>>> CIRCUIT statistics? Who creates this new version?
>>>> I don't understand your question.

What I mean is if there are restrictions on the supported formulas or in the
information gathering process in the PA FM and so on

>>>> The formulae currently supported are specific to the supported entities
>>>> even though they sound generic. For instance, Packet rate, utilization
>>>> Overhead, PDUs forwarded etc apply to Phase4, DECnet/OSI, SNMP entities.
>>>> But the counters used in the calculations are entity specific and their
>>>> semantics might differ from one entity to another. So the formulae to 
>>>> calculate these statistics will be different. This is one thing that will
>>>> vary from entity to entity.
>>>> The actual process of collecting data & calculating statistics will not
>>>> change. 
>>>> The current version of PA is not "generic".

Thanks a lot for your help

Luis 

Guess I didn't explain myself well enough. I mean *we* are planning to develop
a *new* "PA" FM to calculate statistics for a few new entities, obviously
similar to the *actual* PA FM (included in DECmcc BMS).

But one of our new entities is "virtual". It will be used to "group logically"
other entities. I know this is not a formal description for it, but it is
rather difficult to do it without going into the details. To save time just
think of it as "similar" to the CIRCUIT entity in V1.2. That is, it has
attributes that reference other entities.

This is why I asked *what if* the current PA FM had to support statistics for
the CIRCUIT entities (just a mind's game, no requirements). Could it do it by
consolidating counters of the CIRCUIT lines (that are subentities of NODE4 or
NODE)?

Hope this clarifies

Luis

I.  Glad you mentioned about CIRCUIT AM. Being an AM to manage a generic 
    NMF circuits, the Circuit AM in V1.2 does not even attempt to provide
    statistics. But we do have some ideas in mind. Let me toss a few here
    hoping that are useful to you.
    
    1. If the endpoints are same type of entities (example: Node4 Circuit
       to Node4 Circuit), then there it is straight forward to compute the 
       statistics. Here you have the option of taking statistics at either
       endpoint and reporting it. If you plan to take the statistics from
       both the endpoints and do some "creative" calculation, care should
       be taken for the time differential between the two samples. 
    
    
    2. If the endpoints are not homogeneous, try to provide the statistics
       based  intersection of the counter attributes of both endpoints.
       This should include utilization of the bandwidth, and the error rate.
    
      
    3. For the Composite circuits, I don't know. Needs lot more thinking.
     
    
II.    As Anwar mentioned, the  current PA supports statistics only for
       specific entities (like NODE, NODE4, and SNMP etc.). Statistics for
       specific entities need specific formulae, and there is no mechanism
       in V1.2 to enroll these statistical formulae.
    
       If you are computing statistics for entities supported by
       the current PA you may want to use or borrow the formulae used by
       the PA so that the customers DO NOT see two different value for the
       same statistic.
    
       If you are computing statitics for entities which are not currently
       supported by the PA, please keep us posted with the entity types and
       the corresponding statistics with the formulae to be used. That will
       help us to enhance the PA as welll as the Circuit AM in the next version.
    
    
    Seems to be more fun to deal with Circuit like entities. Will you
    moving later to deal with the topology of such entites?
    
    
    Jesuraj
    
    
    
    
                                                  

Jesuraj

Thanks a lot for your help, also for the offline. It has been extremely helpful
for us. I'll keep you updated of any progress.

Luis