Information Technology Reference
In-Depth Information
To manage a large number of different but somehow related attributes, we
can organize them in a tree, called the
attribute tree
. The attribute tree is an in-
memory organization of the attributes of the system resources. One could keep
everything in one level for very simple systems. For complex systems, as the
number of attributes increases, placing them in a tree provides a better struc-
tured organization and, depending on the type of queries, better performance.
An example of an attribute tree is shown in Figure 3.
Fig. 3.
An example of an attribute tree
In the attribute tree, there is a unique access path for each attribute from
the root node. We can compare this arrangement to the files and directories
in a filesystem. Directories correspond to attribute access paths, and filenames
correspond to the names of attributes. The content of the file represents the
different state values of that attribute. Unlike a regular filesystem where you
have either directories or files, here attributes are both and can simultaneously
link to children and contain a value. This behavior is similar to that of the ZFS
file-system
3
.
Attributes are considered to be static and created on-the-fly. A new CPU
could be hot-plugged, which would add an entry in the tree, but those entries
cannot be removed (if a CPU is “hot-unplugged”, the entry will simply not be
used anymore). However, their values may be changing often as the system runs.
It is important to note that no state value (attribute value modification) is stored
in the attribute tree. The purpose of the attribute tree is to give a structure to
the resources and attributes. Each attribute in this tree has a pointer to the
state values in another database in which all current and previous values are
stored. For example, the attribute “current process scheduled on CPU 0” may
have different values during the system execution, which are stored in the state
database and not in the attribute tree.
