[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

115.0. "MIR service" by CROW::YANG () Wed May 02 1990 14:07

    Thanks! Jim,
    I have a lot of questions about MIR, I like to ask you one by one until
    I fully understand its usage.
    
    The next one is: My current Rdb database design for our entity
    hierarchy is really non-trivial. It involves a dozen entities which are
    visible to users plus some other entities which are results of
    normalization of relational database. Users create entities through MCC
    director CREATE directive and manage these entities through other
    directives such as SET, DELETE, ENABLE, DISABLE, etc. Some of these
    entities identify system resources such as processes (i.e. one entity
    instance corresponds to a process to do certain services), other
    entities simply identify distributed resources such as application
    servers which may be distributed on remote nodes. The relationships
    among these entities are also non-trivial, some may be 1-n, others may
    be m-n. 
    
    If using Rdb approach, I can design all entities very carefully
    so that I can manage them very efficiently. One example is: in the case
    of node failure, I can efficiently achieve the failover by retrieving
    information from Rdb database. Another example is: using Rdb database,
    I can design kinds of constraints to help data validation very
    effectively so that if users violate the entity rules, any access to
    the database will be rejected.
    
    My question is this: Can MIR provide the same kinds of usage? In other
    words, if I throw away my Rdb design and rely on MIR to do all that I
    described above, would MIR do the same thing for me? One thing that is
    unclear to me is in the runtime, if a process tried to get the entity
    information from MIR, how is it going to do? Does this process must
    issue an MCC call to do this? This simply means that this process must
    link the MCC calling interface into its image, OR the process must
    issue an RPC call to request a MCC server to access MIR for it?
    
    Anyway, I wish I hear from you very soon.  
    
    Thanks!
    
    Yah-Kong, 
T.RTitleUserPersonal
Name
DateLines
115.1use REPLY not WRITE\GOSTE::CALLANDERTue May 08 1990 16:109
    
    
    RE: note 113 and 115 ***
    
    Please use the REPLY keyword instead of WRITE, so that future questions
    on this same topic will remain within a single topic.
    
    moderator
    
115.2TOOK::SWISTJim Swist LKG2-2/T2 DTN 226-7102Tue May 08 1990 17:3019
    I don't understand what you mean by say "with Rdb I can very
    efficiently....".   Compared to what?
    
    The MIR is not a database system.  It currently provides no recovery
    mechanisms, constraints, etc.   (This may change in the future).  Nor
    does it provide program independence from data definition, unless you
    do something like ILV-encode attribute buffers, which is what most MCC
    MIR applications do.
    
    The MCC MIR is a kernel service not related to MCC_CALL.  All you need
    to do is link with the MCC kernel shared image.  The parameter
    identifying the entity is an MCC Abstract Entity Specification (AES),
    but the MIR does no correlation of the codes in the AES with anything
    defined in the MCC dictionary.  However I would observe that  in the
    normal case, entities would have formal MCC definitions (via MSL) since
    you would be using them in FMs, AMs, and elsewhere in the system.