serverless-application-model VS tokio

The complete solution with SAM is available here.

There are many options to deploy Serverless Applications in AWS and one of them is SAM, the Serverless Application Model. I chose to use it here, because it doesn't add too many layers of abstraction between what's being deployed and the code we write and our infrastructure is quite simple.

The solution uses AWS SAM with the global configuration for Lambda functions and the public API you can use to access Apache Zeppelin. The stack deployment provides the URL as an output value.

When you combine this with the AWS Serverless Application Model you can also very easily include your dependencies. Or use a compiled language like golang for your Lambda functions. You simply run sam build before you run the aws cloudformation package and aws cloudformation deploy commands. SAM will build the binary and update the template to point to the newly built binary. Package will then upload it to S3 and replace the local reference to the S3 location. Deploy can then create or update the stack or you can use the CloudFormation integration in CodePipeline.

I'm a big fan of using an Infrastructure as Code (IaC) approach for any project. My go to tools for this are the Servlerless Application Model (SAM) and it's associated CLI (SAM CLI). For more official use cases and for cross platform apps I typically use Terraform.

AWS SAM

Kicking off the tour and not starting a war, but I'm going to be using the Serverless Application Model.

The diagram here is super simple. I'm going to write something a little later that shows how this code could fit into a bigger workflow, but for now, I'm keeping it basic. And yes, that's the SAM Squirrel in there.

Interestingly, AWS CDK and SAM are both explicitly mentioned. While CDK broadly addresses Infrastructure as Code, SAM is highlighted for its role in developing serverless data pipelines - a hugely underrated concept.

Naturally, there are several options available to declare your cloud resources. The options with the most popularity are the CDK, AWS CloudFormation, SST, Serverless framework, Terraform, and AWS SAM. There are others, but when talking about Infrastructure as Code (IaC), these are the ones you hear about most often.

Being able to control nondeterminism is particularly useful for testing and debugging. This allows creating reproducible test environments, as well as discrete-event simulation for faster-than-real-time simulation of time delays. For example, Cardano uses a simulation environment for the IO monad that closely follows core Haskell packages; Sui has a simulator based on madsim that provides an API-compatible replacement for the Tokio runtime and intercepts various POSIX API calls in order to enforce determinism. Both allow running the same code in production as in the simulator for testing.

tokio / hyper / toml / serde / serde_json / json5 / console

tokio - An asynchronous runtime for Rust

3. Tokio

The AWS SDK makes use of the async capabilities in the Tokio library. So when you see async in front of a fn that function is capable of executing asynchronously.

Petar is also looking at implementing concurrency the way it is in Go to have a fully functional virtual machine as it is in the spec. This would likely attract more external contributors to developing the VM. One advantage of Rust is that, with the concurrency model, there is already an extensive library called Tokio which he can use. Petar stresses that this isn’t easy, but he believes it’s achievable, at least as a research topic around determinism and concurrency.

Another thing to point out is that async is a thing in Rust. I'm not going to begin to dive into this paradigm in this article, but know it's handled by the awesome Tokio framework.

So I started by using Tokio, a popular async runtime. The docs and samples helped me get a simple outbound TCP connection working. The Rust async book also had a lot of good explanations, both practical and digging into the details of what a runtime does.

Regarding the quote:

> The Original Sin of Rust async programming is making it multi-threaded by default. If premature optimization is the root of all evil, this is the mother of all premature optimizations, and it curses all your code with the unholy Send + 'static, or worse yet Send + Sync + 'static, which just kills all the joy of actually writing Rust.

Agree about the melodramatic tone. I also don't think removing the Send + Sync really makes that big a difference. It's the 'static that bothers me the most. I want scoped concurrency. Something like <https://github.com/tokio-rs/tokio/issues/2596>.

Another thing I really hate about Rust async right now is the poor instrumentation. I'm having a production problem at work right now in which some tasks just get stuck. I wish I could do the equivalent of `gdb; thread apply all bt`. Looking forward to <https://github.com/tokio-rs/tokio/issues/5638> landing at least. It exists right now but is experimental and in my experience sometimes panics. I'm actually writing a PR today to at least use the experimental version on SIGTERM to see what's going on, on the theory that if it crashes oh well, we're shutting down anyway.

Neither of these complaints would be addressed by taking away work stealing. In fact, I could keep doing down my list, and taking away work stealing wouldn't really help with much of anything.

The PHP <-> Rust bindings are provided by https://github.com/Nicelocal/ext-php-rs/ (our fork of https://github.com/davidcole1340/ext-php-rs with a bunch of UX improvements :).

php-tokio's integrates the https://revolt.run event loop with the https://tokio.rs event loop; async functionality is provided by the two event loops, in combination with PHP fibers through revolt's suspension API (I could've directly used the PHP Fiber API to provide coroutine suspension, but it was a tad easier with revolt's suspension API (https://revolt.run/fibers), since it also handles the base case of suspension in the main fiber).