How Much Data Does a CI Need?
Raw CIs (the actual devices, software elements, and other real-world entities, not the data stores) contain far more data than you will need in the CMDB. You must trim this overwhelming data set but still maintain visibility into enough to be meaningful and useful.
Consider the example of an SNMP MIB7 for a small branch office router. This device tracks thousands of individual attributes in its internal memory! Try performing an SNMP Walk operation on one, and you will see firsthand the massive set of data that can be gathered. Clearly, you don’t need all of this. Trying to manage such a voluminous data set would be unwieldy, even for a few devices.
Where then do you draw the line to delineate what is tracked and what is not? A good starting point to this answer lies in the DMTF’s CIM standard.8 CIM is an impressive body of work, developed by many brilliant people over several years. It is an elegant and comprehensive object model that is finally getting some traction after a slow start. We go into more detail on the DMTF and the CIM specification in Chapter 6.
The final determination of the CI data’s richness lies in the use cases for the data. Some use cases (for example, fault analysis) require very little detail. Some such as data center automation need much deeper detail. You will struggle to reach the appropriate level of detail on your own, so this is why we recommend starting with CIM. Part of the brilliance of CIM is the work they’ve already put into assessing the right level of detail.