Automatic Profiling ¶

In today's video, we're exploring one of Blackfire's most powerful features: automatic profiling. This feature is designed to automatically enrich the insight on your application performance that are available to you.

Before we dive into automatic profiling, let's quickly recap the three levels of performance insights the deterministic stack of Blackfire offers.

• First, simple traces. Simple traces are high level metrics like overall response time and peak memory usage. These come with almost no overhead.
• Extended traces provide deeper insights with key metrics, such as significant calls and SQL queries.
• Profiles are the most detailed view of your application behavior, from Wall-time to Memory and SQL queries, all the way to the function call level. These provide the richest insights, but come with the highest overhead.

Automatic profiling focuses on collecting profiles automatically, so you can get fine grained insights without manual effort.

Manual profiling is an essential practice for maintaining optimal performance, but has limitations. With automatic profiling, Blackfire enhances observability by automatically capturing a daily snapshot of the 10 most impactful transactions.

It's like having a performance detective working in the background, while you focus on building fantastic applications.

Here's how automatic profiling works. Blackfire analyzes daily data from your production environments to identify the 10 most impactful transactions.

This list is based on Blackfire monitoring data, ensuring the most significant performance outposts are targeted.

The Blackfire agent retrieves this list and shares it with the Blackfire probe. The probe then monitors incoming requests and automatically triggers a full profile when a match is found. One automatic profile is triggered daily for each of the 10 most impactful transactions.

These profiles are captured on real end-user requests to profile real life context.

Keep in mind that profiles are only generated for programmatically named transactions. Unnamed or UI- named transaction will not be automatically profiled

For more information, check the documentation on how to name your transactions.

Now a quick note. Profiling does add a small amount of overhead, so it's essential to use this feature thoughtfully. Automatic profiling triggers just 10 daily profiles, minimizing the impact while gathering valuable data.

When a profile is generated through automatic profiling, you might notice something called pre transaction detection in the call graph and timeline. This happens because Blackfire needs to process part of the request before deciding whether it should be profiled or not.

The detection phase ensures only the relevant portion of the transaction is analyzed, leaving the earlier code execution in a pre-transaction detection node.

To recap, automatic profiling helps you identify the most impactful transaction automatically, gather actionable insights for optimization, save time by eliminating the need for manual profiling.

Using automatic profiling alongside manual profiling and synthetic monitoring will allow you to maintain excellent control and insight into your application performance.