You can use predicates as a way to filter the data generated by an instrumentation, allowing you parse performance data into manageable segments. This can make the data you are analyzing easier to interpret.
In this topic:
Predicates are a powerful feature of Cloud Analytics because they are highly configurable and can drastically cut down on time spent analyzing data. The below image is an example of setting up a predicate. You can see that predicates make extensive use of logical operators.
Follow these instructions to build a predicate.
|Predicates are not currently supported by line graphs. The only way you can build a predicate is through a heat map.|
- On the Instrumentation Element panel, right-click the mouse button. This opens the context-sensitive menu for instrumentations.
- Left-click the Predicate by... menu item. This expands the instrumentation panel to include controls for building a predicate.
- From the decomposition menu, select a decomposition that will serve as the basis for your predicate. The decompositions available here are identical to decompositions you can include with a standard instrumentation.
- Click the operator drop-down menu and select "Equal" or "Not Equal." The operator drop-down menu is how you manipulate the logic you build into your predicate.
- In the Element field, enter the element you want to filter against.
- Click the CREATE PREDICATE button.
The predicate you build will appear above the heat map with which you are working and will resemble a standard instrumentation.
Using decompositions and logical operators, you can configure predicates in any one of several different ways. For example, if you create a CPU: thread samples instrumentation decomposed by application name, you can then create a predicate that filters all results for a specific application. The below image demonstrates how you would build a predicate that filters out all results for the tar application.
You can also create a predicate that filters all results except for results from a specific application. In the above image, if you changed the operator to "Not Equal", the predicate you build would filter out any result from any application with a name other than tar.
In addition, you can build as much logic as you like into your predicate by clicking the New Clause button. When adding clauses, you can use "AND" and "OR" as logical operators to further extend the predicate logic. The below image demonstrates how you would build a predicate that filters out any results for applications named tar and results from any application with a name other than bash.