[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Scope of NIM

>>>>> Andrea Westerinen writes:

Andrea> Juergen, Although I agree with "design patterns", I don't
Andrea> agree with your (3) and don't think that design patterns are
Andrea> enough.  For an existence proof of 3, take a look at the
Andrea> models of the enterprise console vendors (CA, HP, Tivoli,
Andrea> etc.).  The whole goal of these models is to pull together a
Andrea> customer's data - to span vendors and problem domains to
Andrea> manage an enterprise. The problem is that although the models
Andrea> are similar, they do have differences and are proprietary.

Do you really believe an information model defined by the IETF will
make the CAs, HPs, Tivolis converge on their data models?

Andrea> Are you saying that a customer forever must integrate their
Andrea> own data across their enterprise and IETF technology domains -
Andrea> that the experts can't work together to organize and position
Andrea> their work relative to each other?  That we can't get common
Andrea> definition of common concepts and services?  IMHO, customers
Andrea> are so excited about DEN since it offers this possibility.

See, this is the split between what customers want and what technology
freaks and device vendors provide. In fact, there are three key
players in the NM business:

- device vendors (usually participating in the IETF, worrying about
  complexity, strong pressure to get implementations out on the
  market, resource constraint environments, management at the end of
  the food chain in their organization)

- management application vendors (pushing their own technology, mainly
  interested in supporting the big device vendors only, outsourcing
  the actual interface to other devices to third parties, sometimes
  pushing their own instrumentation on devices, typically not
  participating in the IETF)

- customers (the frustated loosers in this game, sometimes forced to
  do all the integration themself, usually not participating in the

The IETF management work has always been driven by device vendors and
agent folks. And from the MIB experience, we know very well that only
relatively focussed work is being implemented and actually deployed.
And it has to be available in a timely manner - otherwise most device
vendors will have their own proprietary solutions and there is no need
to converge to a standard (since customers are too lazy to ask for it
and management application vendors mostly don't care).

Yes, people are excited about DEN because it sounds simple. Despite
the fact that I personally believe that LDAP technically is a bad
choice (boy, this statement will make me unpopular), I have doubts
that the customer's interest in DEN will change the way device vendors
and management application vendors do their business.

Actually, I asked a few weeks ago (perhaps a bit too polemic to be
taken serious) for experience reports how other modeling activities
(such as NRIM) improved other standards, products etc. and if it was
successful, how this work was organized to be effective. I think we
need to analyze these efforts seriously before we rush to repeat
things in the IETF. And as you know, I am very open to get these
discussions going. I think they are important. I am not at all against
modeling - even if this may sounds different sometimes. But we need to
figure out what indeed the right scope is, how we can ensure
timeliness, and we need to understand which role such modeling work
can play in a larger context.

(In this sense, it will be interesting to monitor how PCIM actually
affects policy based management, which BTW every vendor already claims
to ship today.)


Juergen Schoenwaelder      Technical University Braunschweig
<schoenw@ibr.cs.tu-bs.de>  Dept. Operating Systems & Computer Networks
Phone: +49 531 391 3289    Bueltenweg 74/75, 38106 Braunschweig, Germany
Fax:   +49 531 391 5936    <URL:http://www.ibr.cs.tu-bs.de/~schoenw/>