Multigraph User Guide

go/browser-perf-multigraph

This guide explains how to effectively search and filter multiple performance graphs within the Browser Performance Infrastructure. While the Explore (/e) view is designed for deep-diving into a single, comprehensive graph, Multigraph (/m) is optimized for viewing and synchronizing multiple metrics simultaneously.

Note: All examples and links in this guide use the Chrome Perf instance (https://chrome-perf.corp.goog/m/), but the concepts and instructions apply to all Performance Dashboard instances.

Table of Contents

Accessing Multigraph

You can access the Multigraph view in two ways:

Note: This is the default view for most dashboard instances.

Multigraph Sidebar Button

Using the Test Picker

The Test Picker is the primary tool for building queries to find specific metrics. It operates on a dynamic key-value hierarchy:

Test Picker Interface

  • Key-Value Selection: The data is categorized by keys (e.g., benchmark, bot, test). When you select a key, you are prompted to select one or more matching values.
  • Configurable Hierarchy: The order of keys (e.g., benchmark, bot, test) is determined by the include_params setting in the instance configuration. This allows the search hierarchy to be customized based on how the data is indexed (see example configuration). Keys not specified in the hierarchy are not set by default, but can be assigned default values via default_param_selections (see example).
  • Dynamic Filtering: Your selections actively filter the dataset. Once you select a value for one key, the options for all subsequent keys are narrowed down to only show values that match your current query.
  • Traces Counter: As you build your query, watch the Traces counter. This number represents the total count of individual time-series lines (traces) that match your current filters.
  • Plotting: You can click the Plot button to render the graph once your selections are well-defined. Narrow down your choices until the Traces counter shows a manageable number to avoid overloading the dashboard.