Re: Call for censensus on path forward

Hi, 

We have submitted to disman WG a clustering/Replication MIB.  This
presents a fault-tolerant approach for fault-monitoring, in the sense
that objects are available even when part of the agent at the network
element is down, or part of the cluster managers are down, or
unreachable. The proposed approach is a 3-tier architecture for
dependable monitoring by replicating managed objects in middle-layer
managers called clusters. 

By browsing the MIB of a crashed agent you may discover what was the traffic
on its interfaces before it crashed, what type of traffic was there,
and the system state including protocols, CPU, memory, etc.

In order to get this benefit, we must pay a price: the extra traffic
required to replicate selected managed objects. It is possible to determine
replication intervals that will cause an extra traffic that fits within 
acceptable bounds. Furthermore, the distributed clustering architecture
avoids bottlenecks in network monitoring.

The aggregation MIB proposal is a very good approach because it
reduces the cost associated with polling. Since our approach queries
agents repeatedly, aggregation MIB is very useful.


Regards, 
Aldri. 

Federal University of Parana'