Performance

Summarized using AI

​Introducing Helix: High-Performance Ruby Made Easy

Godfrey Chan and Yehuda Katz • April 25, 2017 • Phoenix, AZ

In the RailsConf 2017 sponsored talk 'Introducing Helix: High-Performance Ruby Made Easy,' Godfrey Chan and Yehuda Katz explore Helix, an open-source toolkit for writing native Ruby extensions in Rust. The speakers explain that while Ruby is generally efficient for I/O-bound tasks, it can struggle with CPU-bound workloads, motivating the need for optimizing performance. Key points discussed in the presentation include:

  • Background Context: The original Skylight agent was developed in Ruby, which provided a boost in productivity but faced limitations in performance for certain features, especially when tracking memory allocations.
  • The Promise of C Libraries: Historically, creating performance gains in Ruby often involved writing C extensions, but this posed challenges like increased complexity and maintenance burdens.
  • Rust's Promise: After experimenting with Rust, the speakers found it to be a compelling alternative. Rust offers C-like performance while enhancing safety and usability through a sophisticated type system and guarantees that help avoid common pitfalls seen in C programming.
  • Helix Tooling: Helix simplifies the use of Rust in Ruby applications by reducing the boilerplate code required. Previous versions of the tool faced usability challenges, but improvements have been made, including removing boilerplate while enabling seamless integration into Rails apps.
  • Demonstration: During the talk, a live demonstration showcased how to build a simple Rails application that leverages Helix for CPU-intensive tasks by implementing a method to flip a string using Rust. The integration was facilitated through a straightforward API that Ruby developers could easily adopt.
  • Key Features and Use Cases: The speakers highlighted various use cases like background jobs, mailers, and data processing tasks where Helix could significantly improve performance. They stressed that Helix is well-suited for operations that can benefit from Rust’s efficiency and performance, particularly in computation-heavy scenarios.
  • Future Development: The ongoing evolution of Helix is aimed at broadening its capabilities, including better support for Ruby constructs like optional arguments, more Ruby-like APIs, and enhancing documentation to support developers in the community.

In conclusion, Helix stands as a promising solution for Ruby developers looking to optimize performance for CPU-bound tasks while maintaining the ease and ergonomic benefits of Ruby. The presenters encourage the community to engage with the Helix project, explore its capabilities, and contribute to its development.

​Introducing Helix: High-Performance Ruby Made Easy
Godfrey Chan and Yehuda Katz • April 25, 2017 • Phoenix, AZ

​Introducing Helix: High-Performance Ruby Made Easy by Godfrey Chan and Yehuda Katz

This is a sponsored talk by Skylight.

We got a good productivity boost by writing the original Skylight agent in Ruby, but over time, we couldn't implement all the desired features with its overhead. Ruby is fast enough... until it isn't.

Introducing Helix — an open-source toolkit for writing native Ruby extensions in Rust, extracted from our Featherweight Agent's DNA. Fast, reliable, productive — pick three. Come join us to find out how you can leverage this power in your Ruby apps and even help make Rails faster!

(This is not a re-run of Godfrey's talk from last year.)

RailsConf 2017

00:00:14.450 Good afternoon, RailsConf '17! We'll be talking about the Helix project today.
00:00:21.300 First, I hope you're having a great day.
00:00:27.150 As you can see, I'm here on the Rails forum — yes, that's what we're doing.
00:00:33.780 The most important thing to remember is to run 'bundle update rails'.
00:00:38.910 And as you can see, we are successful!
00:00:46.289 This is actually a sponsored talk by Skylight. I am required to sell you something, but honestly, I don't feel comfortable just promoting.
00:01:10.320 So, instead of selling you on model stuff, I thought I would highlight some really useful talks.
00:01:16.490 There's another talk today that might be of interest. It’s called '160', about using a single codebase for web and mobile applications.
00:01:29.250 If you have to leave, that’s fine.
00:01:36.210 As for myself, I am Grateful, I am here today to talk about a project called Helix, which I discussed briefly last year.
00:01:52.770 This project is an exciting development because Ruby is fast, but we also know it can be slow.
00:02:10.299 Most of the time, Ruby’s speed is not a big concern; however, there are occasions when performance really does matter.
00:02:17.020 To illustrate, let's look at some CPU-intensive workloads that might present challenges.
00:02:31.000 I've been working on this project, along with my colleague, Yehuda Katz.
00:02:37.840 Together, we are both RailsConf 2017 alumni, and we also work on the Ember.js framework.
00:02:49.930 At Skylight, we are hiring senior engineers, so if you're interested, please come talk to us.
00:03:05.860 Now, let me tell you about one feature we implemented this year called 'grades', which compares your app's performance to that of our other customers.
00:03:17.530 That being said, now let's dive into the Helix project. Last year, I talked about Helix and how we have been working on improving Rails.
00:03:42.159 In brief, Helix is an open-source toolkit designed to make writing native Ruby extensions in Rust easier.
00:04:04.419 We’ve dedicated a lot of time to Helix and are excited to share what we have come up with.
00:04:10.299 Everyone here loves writing Ruby; it’s an amazing language, but we also face performance challenges.
00:04:24.720 For example, Ruby applications are often I/O bound, and developers typically spend a great deal of time waiting for database responses.
00:05:06.699 Even when it doesn’t matter, there are times when performance is essential, especially with CPU-heavy workloads.
00:05:24.009 To address performance, one effective approach is using C extensions.
00:05:30.039 This is how Ruby successfully optimized libraries like Date and Pathname.
00:05:38.740 Years ago, Sam Saffron discovered that the 'blank?' method was a hotspot in Discourse.
00:05:44.140 He proceeded to write a C library called 'fast_blank' that sped up operations dramatically.
00:06:01.630 So why don't we do more of this? What’s stopping us from writing more Rails functionality in C?
00:06:07.330 In a nutshell, while C is fast, it's also quite complex and unsafe.
00:06:12.700 Any small mistake could lead to a segmentation fault, which is something we're all wary of.
00:06:18.340 Most developers prefer to stick with pure Ruby to avoid these risks.
00:06:24.850 At Skylight, we faced similar challenges.
00:06:30.070 We found that while our first version of the Skylight agent worked fine in Ruby, we struggled to add a lot of the features we wanted due to performance overhead.
00:07:11.260 Originally, we thought we could solve our performance issues with C or C++ extensions, but we encountered similar burdens.
00:07:23.590 This led us to consider trying Rust.
00:07:29.620 The first trials went so well that we eventually migrated much of our core functionality to Rust.
00:07:38.140 Rust is a compiled, statically typed language that is exceptionally fast.
00:07:49.580 What makes Rust particularly appealing is its advanced type system and ownership model, which guarantee memory safety without a garbage collector.
00:08:10.420 This aspect allows you to avoid data races, making it suitable for concurrent programming.
00:08:25.430 High-level programming languages such as Ruby often present a trade-off between abstraction and performance.
00:08:52.090 However, in Rust, you can frequently use high-level abstractions without paying the overhead cost.
00:09:07.900 The Rust compiler is capable of optimizing code efficiently, often yielding better performance than manually crafted loops.
00:09:35.500 To illustrate this, when we ported the fast_blank example to Rust, we achieved similar performance as the C version with a simple one-liner.
00:09:56.229 While the original implementation required about 50 lines of boilerplate code, we have created Helix to eliminate the unnecessary boilerplate, allowing developers to write cleaner, more concise Rust code.
00:10:22.269 In the past, there was a clear distinction between scripting languages and systems languages, with the former primarily handling I/O operations.
00:10:39.610 However, Helix allows you to harness the power of Rust while writing Ruby, avoiding the need to compromise on performance.
00:11:02.649 Essentially, you can begin with Ruby and transition your CPU-bound logic to Helix as needed.
00:11:21.030 So, to summarize my talk from last year, we’ve made substantial progress in making Helix accessible to everyday developers.
00:11:32.110 We’ve streamlined the setup process and minimized boilerplate code significantly.
00:11:39.610 We are focusing on making it feasible for developers to incorporate Rust code into their existing Rails applications, targeting CPU-intensive tasks to optimize runtime performance.
00:12:09.480 Now, let me show you a demo of a very simple Rails app that flips text upside down using our Rust functionality.
00:12:24.140 We are going to generate a Rails app that allows text input, and upon clicking a button, it will display the reversed text.
00:12:51.860 The implementation relies on our previously defined Rust methods for heavy text transformations.
00:13:17.430 Furthermore, I will show you how seamless it is to deploy Rust alongside Ruby in a Rails application.
00:13:29.990 As we start by generating our Rails app, we're using the latest release candidate, and we're skipping Active Record to keep things simple.
00:13:51.349 Now that our Rails server is running, we will add the Helix gem to our Gemfile and then install it.
00:14:09.090 From there, we use Helix's generator to create the necessary Rust files.
00:14:19.200 The generated files will include both Ruby and Rust components, structured to help streamline your development process.
00:14:35.700 Next, I will implement the core functionality in Rust to ensure efficient processing of the text flip.
00:14:56.960 We will define a Rust method that performs the reversal of the input text.
00:15:06.450 Now that we can compile this Rust code, I will run it to see if our implementation works.
00:15:29.990 After running our tests, we will ensure everything functions correctly before deployment.
00:15:39.490 The Helix project has allowed us to write concise and efficient code, seamlessly combining Rust's performance with Ruby's ergonomics.
00:15:54.490 As we continue to build on this framework, we aim to make the integration as smooth as possible for developers.
00:16:09.200 After setting everything up, we will deploy this app on Heroku.
00:16:33.000 This deployment process is straightforward. You just push your code to Heroku's Git repository.
00:16:46.170 Our app is now up and running, with the Rust code successfully integrated into the Ruby on Rails framework.
00:17:05.690 As for the use cases for Helix, it is particularly suitable for CPU-intensive tasks.
00:17:20.520 Heavy computational problems with simple input types can benefit greatly from Rust's performance.
00:17:40.460 Additionally, integrating existing Rust libraries into Ruby applications is another strong advantage.
00:18:01.900 Helix enables a user-friendly interface to access high-performance libraries written in Rust.
00:18:27.620 Using Helix, you can build background jobs or perform CPU-intensive processing in Rails.
00:18:48.300 We'll also provide detailed documentation on our roadmaps and features as they develop.
00:19:05.490 We invite you all to check out our resources and consider contributing.
00:19:18.230 We are looking forward to engaging with both new and experienced developers interested in exploring the potential of Helix.
00:19:33.940 That's all from me for now. Thank you for your attention!
00:19:56.860 I hope you enjoy the rest of RailsConf.
Explore all talks recorded at RailsConf 2017
+109