[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference azur::mcc

Title:DECmcc user notes file. Does not replace IPMT.
Notice:Use IPMT for problems. Newsletter location in note 6187
Moderator:TAEC::BEROUD
Created:Mon Aug 21 1989
Last Modified:Wed Jun 04 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:6497
Total number of notes:27359

2180.0. "Distributed Directors vs. DNS?" by TROU47::SLEE () Thu Jan 23 1992 13:54

	Version 2.0 will possibly be introducing distributed 
	directors (DDs). Not knowing exactly what this implementation 
	will be, I'm curious to know if and how it may impact
	the use of a namespace.

	Will (DDs) mean that a "common" MIR will be used?
	or,
	Will it be more like any Director will be able to access
	any MIR?

	Will a Director on some platform be able to cause a Management
	Module on some other platform to carry out a directive?
	or,
	Will this Director cause a Director on some other platform to
	have one of its MMs perform the directive?

	As an example, would I be able to configure my 4 platforms
	such that each one would be collecting historical data on 
	individual groups of entities. Then at some interval export
	this data from each platform to a central platform for generating
	reports from the Rdb/SQL file(s).


	If Distributed Directors will mean something like this, then 
	I'm not sure that the advantages of DECmcc using a namespace
	weigh a heavily.


/Steve
T.RTitleUserPersonal
Name
DateLines
2180.1Distributed Directors *and* DNSTOOK::STRUTTManagement - the one word oxymoronFri Jan 24 1992 11:3460
>	Will (DDs) mean that a "common" MIR will be used?
>	or,
>	Will it be more like any Director will be able to access
>	any MIR?
    I think we need to distinguish different parts of the MIR, cos there
    are different answers for each:
    	Class data - initially this will still be stored on a per-system
    		basis, though some day in the future we may find a way to
    		store the information network-wide, yet still perform well
    		enough to make it usable by management modules
    	Instance data - this will still be stored using a Naming Service
    		(DNS or CDS initially). This is to ensure that there is a
    		common global view of the entities used by all directors.
    		[There are those that seem to think that a local MIR will	
    		 suffice - horsefeathers. If the configuration is large
    		 enough to warrant distributed directors, then a real Name
    		 Service is indicated!]
    	Historical data - this will still be stored on a per-domain basis.
    		However, we plan to arrange that the data will be accessible
    		from any MCC.  Thus for all intents and purposes, the data
    		will appear to be globally available - you just need to know
    		the domain.
    
>	Will a Director on some platform be able to cause a Management
>	Module on some other platform to carry out a directive?
>	or,
>	Will this Director cause a Director on some other platform to
>	have one of its MMs perform the directive?
    The difference between the two cases is too subtle for me to
    understand. However, if you look at the SRM (if you dare :-}) you'll
    find a description of what we are doing in chapter 2.  In the current
    implementation, when one management module requests a service from
    another management module, those modules need to reside on the same
    machine.  All that we are doing in v2 is to allow the MMs to reside on
    different machines - and of course, we'll have to do the necessary
    additional magic to ensure that the MMs know how to find each other :-)
    
    
>	As an example, would I be able to configure my 4 platforms
>	such that each one would be collecting historical data on 
>	individual groups of entities. Then at some interval export
>	this data from each platform to a central platform for generating
>	reports from the Rdb/SQL file(s).
    Unless I'm missing something, you can do this today with v1.1.
    You can use the historian on multiple MCC's each storing data into
    their own Historical data store.  Then later on you can export from
    each MCC's store into a single relational database.
    Now what you *will* be able to do with distributed MCC, is initiate the
    exporting requests from a single machine.
    
>	If Distributed Directors will mean something like this, then 
>	I'm not sure that the advantages of DECmcc using a namespace
>	weigh a heavily.
    I don't understand this comment. Perhaps in light of the above
    information you could restate your concern?
    
    Colin