Great article. Very thorough. I'm really looking forward to Hyper (and in turn, Iron, or whatever replaces it) being rebuilt on futures-rs and Tokio. I'm eager to start using this approach in my HTTP programs.
Yeah. It looks like it's under progress, I'm really excited for HTTP futures as well. It will make it exceptionally easy to build highly performant web servers.
It would be so great if Sean got to work full time on hyper, but of course I understand that he is needed elsewhere at Mozilla and he's doing a great job with the time he has. I would hazard a guess that there are so many besides me waiting for the upcoming HTTP client/server improvements that it would probably not be the worst idea if Mozilla (or some other org betting on Rust) threw some money at it.
I'm extremely grateful to everyone working on Rust and its ecosystem (paid or otherwise) for the work that you do. It sure sounds cheesy, but after trying lots of stuff over the years, Rust feels like coming home. There, I said it. I'm pushing it at my employer, and for every piece of tooling or important crate that matures, it gets easier to evangelize.
Yeah, I hear you. That said, money is being thrown, but at tokio itself, rather than at hyper. You have to finish the lower bits before the higher ones.
I'm in a lucky position to be able to choose the tools I use at work. For certain parts, Rust's been already a great help and right now I wait tokio and hyper/tokio more than I waited for Super Nintendo when I was a child. They will be an enormous help when refactoring the current services. I'm pushing myself to learn more so I could be helping with these projects.
That said, thank you for the whole Rust team. It hits the sweet spot being fast, not eating so much memory, being pleasure to write and having amazing tooling around it.
Sean (or other Hyper maintainers), if you're reading this, it'd be really useful to have a tracking issue for the Tokio integration rather than just a branch. I'd like to be able to subscribe to something to keep track of progress and discussion on the topic!
pub trait Future {
type Item;
type Error;
fn poll(&mut self) -> Poll<Self::Item, Self::Error>;
}
... and a bunch of more interesting methods built on top of poll. Futures, as you can see, can be generic over the kind of value they produce, as well as possible errors while producing said value. poll drives the future forward, and should never block. Usually, you don't call poll directly, you create a Task, which represents a chain of futures, and tell it to run. It will then handle doing the right thing, calling the right methods in the right ways.
The second component is mio. Mio is a low-level, asynchronous I/O library, that gives you the standard event loop stuff. In other words, it's a very thin wrapper around epoll/kqueue, and has an adapter to make iocp fit. There are deeper reasons that readiness was chosen over completion as a model, but I won't get into that right now.
So, the lowest layer of tokio proper is tokio-core, which combines futures and mio to give you the ability to say "give me an event loop. Chain some futures together. Run this chain on the event loop." But that's still a fairly low-level interface. And it's what's being shown off here.
At the same level, tokio-service is what gives you the Service abstraction.
At the sort of middle layer, there's a bunch of libraries that you can use, like tokio-proto, which gives a slightly higher-level interface for implementing network protocols.
Finally, the unreleased tokio package combines this ecosystem into an easy to use way to build servers, it's the whole package. This "tons of tiny packages" approach means that if you want to extend tokio in some way, you pick the appropriate level of the stack for your task, and plug it in, and everything up the chain can benefit. It's very modular and extensible.
The key enabler here is Rust's zero-cost abstractions: last time we measured, tokio-core had a very small (less than half a percent, IIRC) overhead compared to writing a mio event loop by hand. And that's before significant profiling effort has been done. So while in many languages, all of these layers would add up to significantly reduced speed, the idea here is that in Rust, they won't. A significant portion of this is zero-allocation or single-allocation, for example.
Part of why I didn't go into it is because the post was already long, another part is because frankly, it's not my area of expertise and I didn't want to misrepresent it. Please remember that I'm not directly involved in tokio's development, and so this is always my understanding, I might be wrong in places, and you should ask someone on the team to be 100% sure. (The post above is stuff I'm sure about.)
What I will say for sure is that it wasn't "hey we use Unix so we're doing what we know and we'll tack on Windows support", it was made purely on technical merits of the two models. I also know that it was related to allocations.
At a high level, my understanding is that the situation is this: the unices have a rediness model, windows has a completion model. If you want a cross-platform abstraction, you can't get around needing to map one to the other. So the question becomes "is it cheaper to simulate readiness with completion, or completion with readiness."
I _believe_ that the situation is something like this: the completion model requires that you allocate a buffer up front, whereas the readiness model doesn't. So to map readiness to completion, you'd end up allocating the buffer, then doing your calls, then filling the buffer. But to map completion to readiness, you can do a trick: make a call for a zero-byte read, and when that comes back, allocate the buffer, and make another call. There's still a small amount of overhead here, but it's less than the other direction.
Again, I might be wrong here.
oh also, shout out to wio, which is mio, but focusing on just windows: https://github.com/retep998/wio-rs I haven't kept up with it as much as this stuff, but I like the idea.
The thing I don't get is why Rust sockets have a `set_nonblocking()` function at all. Should there just be separate`recv_blocking()` and `recv_nonblocking()` functions? It would be much simpler. Sadly you can't even do your own interface like that because there is no `get_nonblocking()` function.
In Unix you don't get a choice to block on one or the other. You just get a non-blocking File Descriptor.
To add this you'd need to iron bandage system calls during compilation depending on which calls were made, and their order they were made.
The alternative is having an Epoll/Kqueue loop run until one or another happen. But that's still a lot of magic and resource consumption behind the scenes for a language that offers low level control.
