New tools for analyzing Android, Linux, and Chromium browser performance – Microsoft-Performance-Tools for Linux-Android

Ivan Berg

Introduction

We wanted to announce and introduce you to a new set of powerful open-source tools for analyzing system performance – Microsoft-Performance-Tools for Linux-Android. Recently, we just released version 1.2 adding Perfetto support, which we will cover in a separate blog post!

Microsoft-Performance-Tools for Linux-Android is a set of trace processing tooling that delivers many of the great features, power, methodology and ecosystem, that has been used to drive Windows performance for the past 15+ years with an evolving and refreshed toolset.

This type of tracing & tooling gives amazing depth & insight into what the OS and app code is doing; and can be used in a variety of diagnostic & performance scenarios. If the analysis scenario also involves Windows, then traces can optionally be co-analyzed on the same timeline with Windows performance traces.

The toolset is built on open-source .NET Core and the new microsoft-performance-toolkit-sdk. This means the toolset is cross-platform and traces can be opened and analyzed on any OS that .NET Core supports! We have worked with the OSS community to build on top of the existing best-in-class open-source tracing for any given OS.

Tracing

For Linux, we support LTTng, perf, and Perfetto. For Android and Chromium browsers such as Chrome & Edge Chromium, we support Perfetto.

Here is the current tracing supported:

  • LTTng (Linux Kernel CPU scheduling, Processes, Threads, Block IO/Disk, Syscalls, File events, etc)
  • perf Linux CPU Sampling(cpu-clock)
  • Perfetto Android & Chromium (CPU Scheduling, CPU Sampling, CPU Frequency, FTrace, Android Logs, Generic Events / Default Tracks, GPU Counters)

The tooling supports parsing & analyzing the above tracing. Capturing & recording a trace is different for each type of tracing tool and is covered in How to capture a trace or logs.

You can process or dump events at the cmd-line using your own custom program with only a few lines of code. For example, there is a LTTngDriver which uses Microsoft.Performance.Toolkit.Engine to drive programmatic access to the traces in the C# language.

Any program or GUI on any OS can process events with the Microsoft-Performance-Tools SDK & Linux-Android plugins. Additionally, if you are optionally on a Windows machine to analyze the traces, the toolkit has great integration with Windows Performance Analyzer so that you can use a powerful GUI to analyze the traces.

Graphical user interface, table Description automatically generated

Figure 1 – Linux LTTng system trace loaded in WPA

Conclusion

We hope you can use the tooling to analyze & improve performance; whatever your OS, device, or scenario! Internally at Microsoft – we are using the toolset to help gain insight and drive performance on these products:

We also hope that you can join the community or contribute to this tooling on our GitHub page.

0 comments

Comments are closed. Login to edit/delete your existing comments