Information Technology Reference
In-Depth Information
store drawable trace objects (events, states, message arrows and so on). It stores
the objects belong to the same time buckets in the same blocks. This solution
works well for the traces with uniform event distribution. However, in cases
where there exists unbalanced traces, in which some parts have more operations
and events than other parts, the tree will be unbalanced and some blocks will be
larger than others, which results in having different response times for different
trace areas. However, our solution does not fix the size of the time buckets and
creates more blocks for the busy areas which makes the response time almost
the same for different parts of the trace.
3 Modeled State System Architecture
The general idea of the solution is to extract and record incrementally the in-
tervals of different state values for system resources (processes, CPUs, disks,
etc.) from trace events. Tree-based organization is used to store the state values.
Figure 2 depicts the architecture of the system which shows the different compo-
nents and their interactions during the construction of the data structure. From
now on we use term “modeled state system” to refer to the architecture shown
in Figure 2.
Fig. 2.
Modeled state system architecture
The modeled state system contains two important parts: the current state
and the state history tree. The current state manages the ongoing state values
for the current time of the trace, while the state history tree encompasses the
past state values of the system attributes. The following sections describe the
different components of the system.
