The MIB is defined as an abstract entity consisting of simple type objects. Simple type objects are objects that have values that are strings or strings that can be interpreted as integers. Simple type objects do not represent an aggregate such as a structure.

Aggregate objects, such as a C language structure, are represented in the MIB by having a separate object type for each element. This can be compared to having a single table for each column in a relational database, with the entire table made visible with a join of all the column tables.

In the model each object has:

Each data object in the MIB has a name, called the classid, which defines its type and semantics. A classid is a string that forms a hierarchical name space by using a period (.) to separate levels of the tree. For example:

In this example, the object is identified as an experimental object (exp). In the current release, all objects are under the experimental branch of the name space tree, owned by the session control facility (scf), in scs code block, and is the scb_query object.

Each data object is located in a place, where the general format of a place string is:

If the place is visible in several different processes in an installation, then it is "vnode specific."

If the place is a particular server, then it is "process specific."

The place can be a local or remote installation. To connect to the remote installation, an installation password must have been defined.

A classid can have none, one, or more instances in a process or a vnode.

Instances have simple data types: strings or strings that can be interpreted as integers.

An IMA table is modeled by having objects with instance values that form a key index.