Overview

This chapter explains how to visualize trace data as you expect. Gallery section shows which type of visualization is supported by CARET. CARET serves uniform visualization API to create such graphs.

Uniform API design

CARET serves Plot class to visualize trace data. The following sample code shows the basic usage of Plot.

from caret_analyze import Application, Architecture, Lttng
from caret_analyze.plot import Plot
from bokeh.plotting import output_notebook, figure, show
output_notebook()

# Load recorded data
lttng = Lttng('/path/to/trace_data')
# Load an Architecture object
arch = Architecture('yaml', '/path/to/architecture_file')
# Map the architecture object to the recorded data
app = Application(arch, lttng)

# Focus on a target callback
callback = app.get_callback('/target/callback/name')

# Get plot object for visualizing callback frequency
# Plot.create_[Metrics]_[GraphType]_plot(data)
# is format to get a target data set to visualize
plot = Plot.create_frequency_timeseries_plot(callback)

# Assign a table to callback_df object
callback_df = plot.to_dataframe()

# Create a graph for frequency of callback execution
plot.show()

plot object is obtained from Plot.create_[Metrics]_[GraphType]_plot(data). The argument of data is, for example, a CallbackBase-based object or a Communication-based object. A list of CallbackBase or Communication is also acceptable as explained later. Any of performance metrics such as latency, frequency, or period is given as Metrics. GraphType is served to select a graph type such as time-series or histogram.

plot object has two method; to_dataframe() and show(). to_dataframe() method returns a table including time-series data on a given metrics. show() method creates a figure of time-series graph and returns the corresponding figure handler. In other words, show() method visualizes the time-series data. If you want to create another type of graph manually, you will get the table by to_dataframe() method and convert it into the expected graph.

Visualization API

This section lists methods to visualize several metrics. You will find a sample figure corresponding to metrics if you access the link. Some of methods are not designed according to uniform API design, and they are exception.

Callback

• create_frequency_timeseries_plot(callbacks: Collection[CallbackBase])
• create_period_timeseries_plot(callbacks: Collection[CallbackBase])
• create_latency_timeseries_plot(callbacks: Collection[CallbackBase])
• callback_sched
  • Visualize callback scheduling

Communication

• create_frequency_timeseries_plot(communications: Collection[Communication])
• create_period_timeseries_plot(data: [Communication])
• create_latency_timeseries_plot(data: [Communication])

Here, CARET takes into account communication when both transmission and reception on a message are performed successfully without being lost. See Premise of communication for more details. Feel free to skip this page if you are not interested in.

Path

• message_flow
  • Visualize message flow of a target path
• create_response_time_histogram_plot
• chain_latency

Helper APIs

CARET provides some APIs which can help users to focus on their respective interest.

• LTTngEventFilter()
• Wildcards for get_callbacks()

Info

CARET analyze API document describes the parameters and returns of APIs.