Conference azur::mcc

5328.0. "DECmcc Distribution News" by TOOK::CAREY () Mon Jul 12 1993 10:49

   
   A little news from the home front:
  
   As of two PM today (friday 9-july) a distributed DECmcc
   management station has received over seventy thousand notifications
   and is still going strong.

   Outstanding achievement!  The setup involves four different
   DECstations and five different DECmcc directors cooperating in a
   single management configuration on ULTRIX.

   The current test configuration is a standard notification and
   alarms stress test transparently distributed over the five directors
   using domain distribution, entity instance distribution and
   notification rollup.  The demonstration is indistinguishable from a
   single station, except that much more CPU horsepower is available.
   
   The setup is simple, there is one new piece of information that can
   be registered, and DECmcc directors all have unique names, and map
   each other through the DECmcc Instance Repository (either DNS or
   Local MIR).
   
   When a domain is created, it can be created using a "Managing
   Director" argument.  If this is specified, it is registered with
   the domain, and whenever that domain is referenced, the call is
   diverted to the DECmcc Director responsible for honoring requests
   to that domain.  Using that, we get easy geographical distribution
   of requests, and can localize network loads close to the place
   where the monitoring needs to be done.
   
   Same thing when an entity instance (such as an SNMP node) is
   registered; specify a "Managing Director" on the registration, and
   whenever that particular instance is referenced, the call is
   automatically diverted to the DECmcc Director responsible for that.
   This is especially useful with devices which can only be managed
   from a single local management station; all calls are diverted to
   that station to be issued, and the results are sent back to the
   DECmcc director that originated the call.
   
   Notification information can be rolled up along the same paths,
   resulting in distribution of the monitoring of the network with
   consolidation of the results.
   
   To the user, the effect is a single management configuration that
   uses many CPUs and locations to accomplish its job.  Management
   Modules don't see the distributed nature of the requests and so
   require only one change (the MSL for registration needs to change
   to include the Managing Director argument).
   
   There is a lot of work yet to do, but getting a stable setup
   running notifications for this long is a significant acheivement.
   The previous "record" was just over two thousand notifications
   before encountering problems.
   
   Just a note to let you  know that there are some exciting things
   happening in DECmcc engineering!
   
   -Jim Carey
   

    Great news Jim! 
    
    Thanks for sharing some good news at a time when a lot
    of us are frustrated with Polycenter et al...
    
    Is there any documentation available on distributed MCC? What protocol
    is used for the director-to-director communication? I know it's early
    to get any concrete info but anything is appreciated... (if you'd
    rather not post here, please send mail...).
    
    Thanks,
    
    Ed

Most exciting!

What distribution technology is being used?  Is it on a migration path
to DME?  Is it particular to VMS or Ultrix or OSF?

When approx. can we expect to see this in external field test, and into 
product announcement?

Does HP Openview have plans for this?  What timeframe?  We're always interested
in differentiating MCC from Openview.

    
    
    	Hurray! Hurray! Hurray!  The light at the end of the tunnel just
    got turned back on.  This was number one on my list and therefore am
    greatly encouraged.  
    
    	While I know this functionality is not available for sale both Ed
    and I are scoping/crafting network management solutions for large WAN
    environments ie. greater than 1000 sites.  We could use what ever
    information is available regarding the mechanics of the distributed
    environment so that we can take this into account in our designs.
    
    	I'm willing to sign papers, buy dinners, order drinks, whatever it
    takes.
    
    Scott

    What about mixing VMS and ULTRIX(OSF) in one Environment ???
    
    In Austria we have the (maybe special) Situation that 95% of all
    MCC Customers run commercial VMS Applications and want MCC on VMS.
    (familiar environment, VMS Mgt Tools etc...)
    
    But it gets increasingly difficult to stay with VMS alone
    => Launched Mgmt Applications, Cable Manager etc. want UNIX
    
    So maybe Distribution makes it possible to use both Platforms in a
    integrated Way (without much additional work...)
    
    Is this a possible approach ???
    
    					Guenter (also seeing Light now...)
    
    
    

    	any news on this ?
    
    	danny

    We're hoping to deliver the Distribution functionality to TeMIP
    development team in France.
    
    -Matt.