RustConf 2023 - Profiling async applications in Rust

Rust

Profiling async Rust apps can be complex. Learn how `diagnose` and tracing approaches provide accurate performance information and overcome profiling challenges.

Key takeaways

Profiling asynchronous Rust applications is complex due to the non-deterministic nature of asynchronous code.
The diagnose macro can be used to wrap futures and provide profiling information.
Traditional sampling profilers are not suitable for asynchronous code as they rely on sampling the execution of a program at specific points in time, which is not feasible in asynchronous code where the order of execution is dynamic.
Using a tracing approach to profile asynchronous code can provide more accurate information about the performance of the application.
Generating stable load for profiling purposes can be challenging and can lead to inconsistent results.
51% of the main function time is spent in the extend function, while 49% is spent in the string concatenation.
Buffer extension is faster than string concatenation in this particular example.
Interactive profiling can be difficult due to the dynamic nature of asynchronous code and the need to handle a large number of threads and concurrent tasks.
The Tokyo runtime can hide information about the execution of the program, making it more difficult to profile.
Tracing can be used to profile asynchronous code and provide information about the performance of the application.
Flame graphs can be used to visualize the execution of the program and identify performance bottlenecks.
Interactive profiling can be challenging due to the need to handle a large number of requests and concurrent tasks.
The diagnose macro can be used to provide profiling information for futures and asynchronous code.
Widely using profiling tools can lead to better tooling and a more mature ecosystem.
Interactive profiling can provide insights into the performance of the application and help identify performance bottlenecks.
The diagnose macro can be used to provide information about the performance of asynchronous code and help identify performance bottlenecks.

RustConf 2023 - Profiling async applications in Rust

More talks