|
>> My group is currently building a transport protocol for wide-area reliable
>> broadcasting. The implementation will be in an Ultrix process.
>> I need to make this implementation DECmcc manageable. By this I mean that there
>> are characteristics and counters that need to be set and queried and that we
>> will need to ability to reliably log events.
>> From what I can gather, there is no problem with this on the Director end but
>> there may be at the "agent" end.
>> Although this is a low-level protocol, I believe from mcc's point of view it
>> is an "application".
You are correct that even though the protocol being implemented by the
software running in the ULTRIX process is low-level, the director treats
it like any other application.
A application can be defined as manageable entity, the same as any other
entity (physical or logical). In order for the director to manage the
application running in the ULTIRX process the application must respond
to some management protocol. This protocol can be one of the standard
ones that the director already knows how to interpret or a protocol
unique to the application. The piece of the application that
communicates with the director is called the `agent'.
The director speaks to the application agent using that management
protocol to effect changes in the behavior of the application or to
monitor the values of attributes of the application. You have multiple
choices:
1) Write an AM that plugs into the director and speaks the
protocol understood by the broadcast application
2) Use the Common Agent (currently under development)
development software to produce an agent for the broadcast
application that speaks a protocol that the director
already knows (SNMP or CMIP).
The tradeoffs are : The AM takes more time to develop, but the CA
software is still under development and will not go to field test until
later this summer. Contact Marian (BLUMON::) Weisenfeld for more info
on CA.
In either case, the level of effort is more than 2 weeks. You must
specify the management piece of the application:
1) Structure of the manageable part of the application
2) What operations are supported? Possible replies to those
(success, failure, returned arguments, datatypes, etc)
3) What properties does the application have (attributes,
events, etc)
This specification can take some time, if you do not already have it in
necessary detail. We find that most engineers specify management
capabilities of entities in much less detail than they tend to specify
the operational aspects of the entities.
In either case (agent or AM mentioned above), you must write routines
in the application that cause the management operation to actually
happen in the entity.
>> What will I have to do to make this protocol implementation manageable? I need
>> to get my hands around this ASAP so I can schedule the activity of making it
>> so. Right now, I can't say if it's a 2 wk effort or a 3 mo effort!
>> Any help/input on this will be greatly appreciated.
>> Al
If you need more info, contact me, Claudia (CLAUDI::)Peters, or Marian
Weisenfeld directly.
...kjn
|