The MPLS WG Archive

["Thomas D. Nadeau" <tnadeau@cisco.com>]

> > >4. How many notifications can be generated per second?
> > 
> > Are you asking for a worst case scenario?
> > 
> > In a network where LDP is configured on each device (and not being
> > reconfigured) would not expect notifications to be generated once 
> > sessions are established.  So would not quantify X 
> notifications per 
> > second.
> > 
> > Maybe another way to ask this question is: how long does 
> the average 
> > LDP Session last?
> > 
> > Would say most last longer than a second, but
> > would be good to get operator experience here.
> > 
> The thing is that we do not want the network and the NMS to be flooded
> (congested) with notifications. If there is a true story that they 
> will be infrequent, then just add that explanation, and we're 
> OK. Do remember that a NMS may be receiveing such 
> notification from many managed devices. So if you had one a 
> second per managed device, and you had to manage 1000s of 
> devices, then there is potential still for NMS 
> overload/congestion. So some words to make people feel 
> comfortable is goodness. If the evaluation is that there 
> might be too many notifications, then we may need to add a 
> throttle object.

	Bert, 

	I don't understand what you are getting at here.
How can any one device know what the other devices in the
network are doing? Given your description above, you seem to
be advocating that every notification have a throttling value
associated with it. Although I totally agree that this is something 
that is often appropriate on a per-box-basis (indeed my boxes do this), 
having to put this into every new notification seems difficult to
implement at best. Imagine if you have accidentally not set 
these values across sevearal MIBs to be inconsistent? Imagine
the fun the agent has to go throught to appropriately throttle
each MIB's notifications.

	--Tom