[Fusioninventory-devel] A better strategy for network inventory, based on SNMP models analysis

Gonéri Le Bouder goneri at rulezlan.org
Sun Jul 28 12:31:57 UTC 2013


On Fri, Jul 26, 2013 at 05:21:33PM +0200, Guillaume Rousse wrote:
> Hello.
Hello Guillaume,

(...) 
> Unfortunatly, we don't have any variable appearing with a constant value in
> all models, that could easily be hardcoded in the agent. But many of those
> variables, despite not always present, are constants (uptime, for instance),
> and many others are almost constants (contact, for instance).
Some models may be also broken.
 
(...) 
> For others, excepted for the most variables or barely present (firmware,
> serial, portDuplex, firmware1, firmware2, otherserial), I'd suggest using an
> hardcoded default value in the agent, which may eventually get overrided by
> a model. This way:
> - the agent could achieve minimal inventory without a model (for instance,
> directly in discovery phase)
> - the agent could retrieve more information for devices for which the
> current model is actually incomplete
> 
> And by removing this now useless redundant information from models:
> - their size would decrease
> - their number would decrease
> - their potential error risk would decrease
A agree, I also think the need to have an up to date model database will be
lease important.

> I'm gonna test this strategy against our existing SNMP walks database.
This is a great idea. For your information, MongoDB¹ may be very useful
to do statistics on a large number of models.

¹this one or some other nonSQL DB.

Best regards,
-- 
    Gonéri
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/fusioninventory-devel/attachments/20130728/e1285b80/attachment.sig>


More information about the Fusioninventory-devel mailing list