Conference azur::mcc

An off line offer of help has been made...

Winton,

You should also work with Pete Savage (KEEL::Savage) who formerly was the MCC technical implementation
leader and now works for Bob Daley as technical leader for the X.400 MAIL/Messaging effort.

Pete plans to provide a MOM for the Common Agent for managing the Mail User Agents.
The Common Agent interface supports User management from DECmcc and other directors.

\Pete

    Quick question....  I'm confused when I'm trying to draw the E-R
    diagram for a Homogenous Cluster...
    
    In a possible configuration, a cluster may have 1 queue (for electronic
    mail), and 0..1 senders on each node in that cluster. As the ER
    guidelines only talk about the node - how do I do this ?
                
    Cluster has 1 queue
    Cluster has 1..n Nodes 
    Node has 0..1 sender servers
    queue has 0..Servers.
    
    But I thought that the NODE was the basic entity - so how can I 
    model this ? 
         
         

    A further contact is the Mail Transfer Agent people in Reading England.
    They added MTA as a child entity of the DNA5 node. 
    
    Also, If you would look into what is available via DNA5 node you would
    find that all 7 layers of the OSI reference stack are supported as
    of V1.2. The have added additional child entities that include
    OSAK, MTA, etc.
    
    wally

Re .0:

This reply's a bit late, but I can't keep up with the MCC file. Questions by
mail, get answered quicker...

Don't know how to make it clearer - build a MOM for ALL-In-ONE 
for the Common Agent. The notes conf for the Ultrix CA is on 
DECWET::COMMON-AGENT, for VMS its at SYSMGT::DMS-IFT.

If you want a special application (the generic applications don't do
enough for you) then you ought to build a MCC FM or PM (or both) to implement
the application.

Re .3:

Well, a Cluster is just another node. If you have 4 VAXen in a Cluster, 
that makes 5 nodes (4 "cluster members" and 1 "cluster").

If you have a queue that covers the whole cluster, then that is related to
the cluster as a whole. Just for grins, I quick threw together an EER diagram  
for you using a tool called DECdesign (yes it's a product we sell). I'll try and 
post it in the next reply as a DDIF file. If we're lucky, it will pop
up on your screen if you're using DECwindows notes.

The diagram doesn't show the "module" level entities that ought to be 
introduced into the containment hierarchy between the nodes and the Queues and
Servers.

Now there's a couple of important implementation issues that show up
with this design. One is that because the Queue lives on every node in
the cluster (i.e. every member), then a directive sent to that entity
needs to be coordinated with what's going on with the other nodes in the
cluster. Cluster wide locking will help you there. I presume the product would
have to do something like that anyway so two servers don't take the same mail
job off the queue at once. The second is that the VMS Common Agent hasn't yet
been tested with the idea iof a Cluster node. In any case, it will only
deliver a directive to a single MOM running on one of the cluster members,
it's just that you can't tell which member will get it.

There's some discussion of clusters in a white paper I wrote on
distributed entities at least a year ago. Check out
FILES::EMA$:[Public]WP-Distributed-Entities.PS or something like that...

Mark

Well how about that, it worked

    I tried it on a vanilla terminal and the file is just one long line, so
    don't be surprized.
    
    Mark