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

RE: Call for censensus on path forward



At 9/21/2002:08:00 AM, Wijnen, Bert (Bert) wrote:

Hi Bert,

I'll take your request for feedback on this specific
issue as an opportunity to offer some general feedback
as well:

>I would really like to hear from a few operators (or
>even NM application developers) if they indeed find it
>a requirement to do polling a 1-second-granularty.
>It does not sound realistic to me... but who is me?


1.  If a management application requires sub-second
data updates, a publish/subscribe model should be used.
For SNMP, the Inform-PDU can perform this role for a
large set of application scenarios.  Of course, that
implies MIBs (and agents which implement them) which
support the necessary registration process for Inform
clients.

2.  Of the current proposals under discussion, the only
ones that I could support are the "Extended Capabilities
Negotiation" the "Aggregate MIB"...approaches that do not
modify the underlying protocol.  However, both of them
need to be modified for use in an AgentX extensible
environment.  And I don't really see the absolute need
for either of those proposals to achieve what I think
are necessary goals for SNMP at this time.

3.  That last comment leads me to my major point:  I am
not especially enthusiastic about any of these proposals.
My summary view is that we need more capable MIBs -- and
agents which implement them, of course.  The trend in
large-scale service provider network management technology
today is toward more "bottom-up" capabilities via NE software.
Coupled with increased security requirements for that same
software, the renewed (it's a cyclical thing, you know)
focus on bottom-up, presents an opportunity for resurgence
of SNMP.  Protocol stability -- at v3 -- coupled with
more capable MIBs and agents -- is a fundamental requirement
at this stage.

4.  Three final asides:

   a.  I draw the above observations from detailed
       familiarity with a number of major service
       provider procurements currently underway.

   b.  At the risk of showing my age, I believe that
       MIBs can be made *vastly* more capable -- in
       terms of, e.g., aggregate/derived objects,
       behavior/operations, secure sets, multi-entity
       interoperation, publish/subscribe services, etc.
       -- without major mods to the SMI and definitely
       without modifications to SNMP[v3] itself.

   c.  The primary role of management applications in
       the new paradigm will be to translate user policy
       configurations into agent (MIB) configurations,
       let the agent do its work, and inform the user
       of operational exceptions relative to the behavior
       stipulated by the policy.

Cheers,

BobN