The End of Moore's Law & the Need for Concurrency
I remember my first computer was a Pentium 100 megahertz. I would be in school and then a friend in school say 'I got a computer that's like 233 megahertz' and then like 'how! I just bought mine - mine is new and then now you have something that is like twice and a little bit more faster'.
But that's how it happened for a long time. The hardware, every two years, it would get, for example, our CPUs, every two years, it would get twice faster. And for software engineering in general what it meant for programmers, developers, is that if you write a code today and then don't do anything with this code, two years later you would have the opportunity to run this code twice faster.
But this is no longer true today. So what it means that before we had only one CPU that was getting faster and faster, but now it needs to have multiple of them. And that actually changes the way we write software and the languages that we can see, the mainstream languages, they are not as efficient, as effective as it could be.
Imagine that, you know, in the '80s, the telephones, it's really getting widespread, everybody's installing a phone. If I have a city with millions of people, right, you need to install a bunch of telephone switches and you have calls coming all the time, right, and calls going out, so information is coming and going. When it's working hours I have much more people using the phones, right, and all this kind of stuff. So they have to solve this problem three decades ago - and they did!
So what happened is that Ericsson, they created Erlang. This technology that they created, right, in the eighties, to solve all these problems, it's going to be perfect to solve those issues that we're having right now with concurrency, those issues that we're having with the web in general, right?
I think that was the moment when I had the idea of creating a programming language. Like, 'look I have this absolutely beautiful piece of software which is the Erlang virtual machine. I want to use it more but it's missing some stuff and I want to try adding this missing stuff.' So in 2012, it was the moment that I started working on Elixir part-time and I think we can use the software to expose an alternative for the programming community if they are interested in solving all those challenges that we're having right now.
Solving Complex Aggregation Challenges at FromAtoB
You can do everything on the internet. You can find everything, you can book everything, but it's still in silos. So you will find your flights on Skyscanner and Kayak; you will find your rail connections on bahn.de. Here you have to figure out: how do I get to the airport? How much will my transition time will be? What is the prices of different tickets from different airports and so on?
And we thought it would be a great idea to have this in one UI. You would enter your direction and you will enter your destination and we find the whole itinerary. We immediately realize that this is something that is very hard to do. For every search that we do we have to open a lot of connections to different transport operators. We have a lot of connections open at the same time to different operators and you have to consume this data and process it in a smart way and it has to be fast. This was basically the point where we thought about what technology is the right technology to go for. One of the characteristics that we were looking for was scalability. We literally have tens of thousands of connections open all the time and in both directions and this is one thing that Elixir is very good for us and very low in terms of the resources you need on the server side. We pretty fast came to the conclusion that Elixir is the right technology for our problem.
Team Building and Infrastructure Stability
We used, of course, job boards like Honeypot and others to find people. We also basically train people that did something else before and train them to Elixir.
Of course we have some hiccups sometimes, in the code or we have some bugs that we created but we never had outages in terms of the infrastructure which is for me a very good indication that this is the right technology decision because I never experienced this before.
Live Demo: How Distributed Elixir Works
So one of the big things about Elixir, and we get exactly from building on top of the Erlang VM, is that we can write distributed software, software that runs in more than one machine. So you can see here I have two machines and so what I'm going to do is that I'm going to start one Elixir session on this machine and another Elixir session on this machine. So in this one I'm starting something called interactive Elixir on both of them but this one, this is Bob's machine, so I'm going to say, hey this is Elixir and this has the name of Bob and we have a secret here; a cookie, which is a secret, but we'll just call it 'secret' right now. So I'm going to start this session here on Bob and I'm going to do the same thing on Alice. So I'm starting Alice's session with the same cookie secret.
All right, the only way to do this is that I'm going to define some Elixir code and I know this is a little bit cliche, right, but I'm going to define a module called 'Hello'. We have a function called 'world' in it. And what all this function does is that it prints 'hello world'.
So I define this code, ignored the whole gibberish for now, but I define this code and now I can call hello.world, and you can see that it prints 'hello world'. Beautiful, right? So, it works. This is very exciting.
So this is Bob's computer but now let's go to Alice's computer. If I try to evoke hello.world, it doesn't work. Why? Because I defined this code, I defined the module Hello with the function world only on Bob's computer, but we can solve this, right, because it's distributed. I can make those machines talk to each other. So what I'm going to do here is that I'm going to say hey I want on Alice's machine, she's going say 'hey, I'm going to tell Bob to execute this code for me'. So I would say, 'hey, Bob', the Bob instance that is running on Bob's machine, 'I want you to execute the hello.world code'. And now when I do this we can see that we got 'hello world' printed back. We can see that it works right because Alice was able to tell Bob 'execute this code for me'. Bob executed that code and said, 'hey, I executed this code and this code has a hello world message that I'm sending back to you so you can print'. And that's it.
And what is really exciting about this is that we can build a bunch of interesting technology that just runs on these nodes being connected. So for example, the Phoenix web framework, it has a presence feature and has a Pub/Sub mechanism. So with the Pub/Sub, what you can do is that you can send messages to everybody connected to any machine and with the presence feature you can know everybody who is connected in the whole cluster. You can know who is joining, who is leaving and we can do all that without adding databases, without adding third party dependencies. We just use the distributed feature which is really exciting.
One of the things that we did since the beginning is that the development was always open; always open source. So anybody could join at any time and give their ideas, contribute, share. So I also knew from the beginning that if I wanted this thing to happen I would have to go out and talk to a lot of people and go to different events and convince developers why they should care about the technology, what are the potentials of the technology and get them excited about it. We really started to see like an uptick in people using Elixir and that was kind of like the beginning of a turning point because people are actually starting to bet on this. So you don't feel alone in the sense that you know there are other people believing in the potential, wanting to push it forward. And you know, the interest, it only started to grow and grow. In July 2014 was when we had the first Elixir event.
Voices from the Community: Culture, Docs, and Use Cases
There's such a great vibe, everybody gets along and it's so easy to be able to just approach anybody that's there to ask questions.
When I started with Elixir, I was amazed by the documentation that even was available at that time.
Elixir is always typically thought of as being great for concurrency and fast and so if you're kind of solving those kind of problems then it's a really good choice. Whether you're building something to run on a Raspberry Pi Zero, a five dollar computer or a forty core server, Elixir is going to be fantastically suited for it.
You can't exactly replicate the environment in which people from the community are all in the single place. Elixir is going to be on a track to start to conquer more and more industries.
The Future of Elixir: From Web Development to IoT
Some of the obvious ones that we've seen so far have been working with web development but the scalability there isn't just confined to building websites. We work together with Chris and Jose from the Nerves side of things to be able to try to increase the number of concurrent connections and push the boundaries of a how many devices we can actually get simultaneously connected to Phoenix, for example, because we really believe that it has capabilities of Internet of Things connectivity on the scale of millions of devices. And having that kind of connectivity starts to really open up the language and the industry to start working in all these different kinds of ways.
Unleashing Scalability with the Phoenix Framework
Phoenix is a web framework for the Elixir programming language and it really is like a batteries-included web framework for the platform. The first version of Phoenix, as written, supported like 30,000 users on one server and then we made like 10 lines of change to the code and that's what gave us 2 million users.
What kind of problems or businesses could I build if what before took 100 servers could today take two servers? And I think that's enabling a lot of innovation. So I think that's what's kind of bringing people in is this promise of things that were either impossible to do before or prohibitively expensive to do. Now as a single developer or a couple people you can come in and build something really compelling that wouldn't be doable before.
Global Growth and the Power of Decentralized Contribution
Right now we are at the Elixir conference in Warsaw. We see Elixir conferences appearing all around the world so we have in Mexico, Brazil, throughout the United States we have a bunch of smaller conferences popping up. And then there are all the different talks where we can learn new things and also which is always very interesting is to learn like use cases. You build the tool and then people started using those tool in these very different ways. They're like, 'look, I actually realized that this tool is also very good for this' and then you're like 'it makes sense, like, I watch the talk and it makes total sense'.
So, all those things are very exciting. If I try to centralize and I try to do everything on my own I won't be able to do it. But if we say, you know, everybody can contribute a small part to this and everybody together, decentralized, can do that and bring the community forward, then we have a chance of actually making a lasting impact.
Beautiful.
