Subscribe to receive notifications of new posts:

Profiling Your Workers with Wrangler

2021-09-18

2 min read

In the year since Cloudflare’s launch of Workers Unbound, developers have unlocked the ability to run computationally intensive workloads on the Cloudflare edge network — like image processing, game logic, and other complex algorithms. With all that additional computing power comes a host of questions around performance. Our customers often ask us how they can profile or monitor their Workers to see where they spend the most CPU time, or to see whether their changes improve performance.

Here at Cloudflare, we not only want to build the fastest, most affordable, and most flexible compute platform at the edge; we also want to make the lives of our developers easier in building their applications. To do this, Cloudflare has begun to integrate with existing tools — places our developers feel comfortable and efficient in their day-to-day work. To help measure performance of our customers’ Workers, we’re beginning to integrate with the Chrome DevTools protocol. Just like you can use chrome://inspect to debug your Node backend, you can also use it to profile your Cloudflare Workers.

Introducing Chrome DevTools Integration (Beta)

We’re starting off this integration with beta support for local CPU profiling, using Wrangler. To show off how to use this feature, I’m going to be optimizing a simple JavaScript program which outputs the first thousand integers separated by a space. Let’s start by installing the latest version of Wrangler. We’ll also need a Worker cloned down to your local machine. I’ll be using Workers Profiling Example, but feel free to use any CPU intensive Worker for this tutorial.

For reference, my sample code is below. You’ll notice this code is intentionally performing a computation that will slow down our Worker.

addEventListener("fetch", event => {
  event.respondWith(handleRequest(event.request));
});

async function handleRequest(request) {
  let body = '';
  for (let i = 0; i < 1000; ++i) {
    body = body + ' ' + i;
  }

  return new Response(body, {
    headers: { 'content-type': 'text/plain' },
  });
}

To confirm this, we’ll run Chrome DevTools and profile our Worker locally. In the directory of your sample project, run wrangler dev --inspect. To launch the DevTools, open chrome://inspect in Chrome or Chromium. You’ll see Chrome’s DevTools homepage pop up:

Open up Chrome DevTools using chrome://inspect in your browser

Click ‘Configure’ and add localhost://9230 as one of the targets. You should see wrangler[{Worker name}] appear under “Remote Targets”, where {Worker name} is the name of your project.

Click ‘Configure’ and add localhost://9230 as a target

Click ‘inspect’, then in the popup, ‘Profiler’.

Once in inspect mode, open up the Profiler

Click ‘Start’ and, in a different browser tab, make a few requests to your site, then click ‘Stop’. It should be available locally on localhost:8787.

Analyze the flame graph. For my Slow Worker, I see the following graph:

Flame graph generated from the recorded profile of our sample Worker

When I click on handleRequest, I see the following annotated source of my Worker:

Click on one of your functions to view the source code

In particular, it shows that, unsurprisingly, most of the time is being spent on memory allocations in the body of the loop. Note that the profiler is not always accurate with regard to exact timing, but does paint a picture of your largest bottlenecks.

Understanding the Profile

The profiler works by gathering a stack trace of your program at a sampled rate, so remember that the profile is an approximation of where your code tends to spend the majority of its execution time, and is not meant to be perfectly accurate. In fact, functions that execute in less than 100 microseconds have a chance to not appear in the profile at all.

What’s Next?

With the Workers platform, we’re striving to build in the observability our users expect out of a robust compute platform. We’d love to hear more from the community about ways we can improve visibility into the code you’re writing on the edge. If you haven’t already, please join the Cloudflare Workers Discord community at https://discord.gg/PX8s2TmJ7s. Happy building!

Cloudflare's connectivity cloud protects entire corporate networks, helps customers build Internet-scale applications efficiently, accelerates any website or Internet application, wards off DDoS attacks, keeps hackers at bay, and can help you on your journey to Zero Trust.

Visit 1.1.1.1 from any device to get started with our free app that makes your Internet faster and safer.

To learn more about our mission to help build a better Internet, start here. If you're looking for a new career direction, check out our open positions.
Speed WeekWranglerCloudflare WorkersDevelopersDeveloper Platform

Follow on X

Kabir Sikand|@kabirsikand
Cloudflare|@cloudflare

Related posts

September 27, 2024 1:00 PM

AI Everywhere with the WAF Rule Builder Assistant, Cloudflare Radar AI Insights, and updated AI bot protection

This year for Cloudflare’s birthday, we’ve extended our AI Assistant capabilities to help you build new WAF rules, added new AI bot & crawler traffic insights to Radar, and given customers new AI bot blocking capabilities...

September 27, 2024 1:00 PM

Our container platform is in production. It has GPUs. Here’s an early look

We’ve been working on something new — a platform for running containers across Cloudflare’s network. We already use it in production, for AI inference and more. Today we want to share an early look at how it’s built, why we built it, and how we use it ourselves. ...