You are planning to build something in 2016, aren’t you? Some decisions have to be made, team to be hired. For a few years now you have been relying on Ruby (and Rails) for all your work and you have been reasonably happy. You have been hearing a lot about this new Elixir thing and considering giving it a shot? Let me tell you a bit more about it and help you make a good decision.
Ruby on Rails
A few years ago, Ruby on Rails exploded with popularity. A lot of people switched from PHP to build web-based systems better and faster. Plenty of Java folks made a switch too, being happy to deliver things way faster now. Many advantages Ruby had over other languages are still valid. The same can be said about Rails framework. The company I am happy to be part of, AmberBit has been building Rails-based apps to satisfy clients for many years now. Just as many other Ruby developers, we have been looking for alternatives for some time now. You probably want to know why.
The web changed a lot since the year 2006. If you don’t believe me, have a look at some of the sites on web.archive.org. The web changed it’s look and feel, but also changed the way it works. The applications are way more heavy on the front-end. They need to speak to different variety of devices, and usually handle more traffic. The overall number of Internet users tripled during last 10 years. Users’ expectations about web apps changed too. People don’t want to wait or reload page to see the effect. They want to see more and more live/instant updates. Almost all systems we have built at AmberBit these last couple of years, had more or less features like that. We have been joking about making a “real-time Rails app” template that integrates Faye, share.js and couple of other useful tools into the standard stack. In fact, this is where Rails slowly evolved in the end by introducing ActionCable.
When to stick to Rails in 2016
Having said all the above, Ruby on Rails may be just enough for you. Let me show you a few cases where I think it’s still a great fit.
1. Internal apps for your organization
If you want to build internal app, that is heavy on forms and user input, Rails may be a perfect choice. Those sorts of applications are usually not receiving so much traffic. They are not exposed to the general traffic on the web. You don’t expect anything fancy. All you need is a tool that will help you improve the processes in the organization. Ruby on Rails comes with good support for data validations via third party or built-in functions. There are libraries to choose for authentication and user management. You can also pick some templates to help you style the thing on your own. From my experience, a lot of companies rely on building such apps. They are willing to spend limited amount of money on building them. The budgets tend to be reasonable - yet limited. Rails helps to meet those constraints.
Not all prototypes I may say. This depends on the system you are building. The nice thing about Rails is that you can put together a “scaffold” for your app in just a few days. User authentication, administrative dashboards, permissions management come all in form of third party plugins. This means your developers don’t need to write all the code to handle the boring stuff. They will spend most of their time building the core features instead. That is what makes your system different, unique. The stuff that investors will look into when making financing decisions.
Word of caution here: you may be required to dump the code base altogether at some point in future. Scaling Rails is not easy, and one of the major reasons you may want to go with Elixir instead!
3. Systems with limited target user base
In many cases, your system will be targeting specific group of people. If you are not expecting hundreds of thousands of concurrent users at the same time, you will avoid the scaling problems altogether. Ruby on Rails does not scale easily I am afraid.
4. Systems that nicely divide into smaller subsystems
One of the patterns we have been using to work around issues with scaling, is splitting the apps. If you can divide your monolith to smaller parts easily - go for it.
Some systems can be split into identical instances serving different people. You can run a number of instances of your app, on dedicated servers instead. It can simplify the design too. Multi-tenancy is not an issue, since apps connect to dedicated databases.
Good example of a system that may be doing that behind the scenes is Slack. You set up organization account, that is identified by a unique URL. Those URLs for may be pointing to the machines running independently. Doing this will also make your systems independent of each other. Being more resistant to total system failures is an advantage. Worst thing that can happen is the failure that affects all your clients. You want to keep the failures as much localized as possible.
5. Systems that can use caching effectively
To avoid scaling issues (again) you may want to use caching techniques. Ruby on Rails may be still a good fit for you if you are expecting high traffic, but you can cache most of your pages. This is very true for web publishing platforms, blogs, news sites and so on. We have been using caching to avoid system overload in times of high traffic spikes in many projects now. That said, not all high-traffic systems can use content caching in equal amount. Caching is hard when your system provides personalized experience.
When would I recommend choosing Elixir over Rails?
1. High-traffic systems
Typical Rails app responds in tens or hundreds of milliseconds to web request. Elixir applications can respond in microseconds instead. This can be put to great use. You can handle more requests on the same hardware. Elixir is simply faster by an order of magnitude.
Erlang and Elxir, are great fit for modern hardware. It has been evolving to having more and more number of processors. Ruby will have issues trying to provide work for all the cores. This is the same case with other “traditional” programming languages - like Python too. Elixir’s programs are built in distributed manner even if they run on single machine. This allows you to use the resources you have better.
2. Systems that do stuff asynchronously
Integrating with external services may require doing things asynchronously. Ruby and Rails is not really great with waiting and responding to asynchronous events. Sometimes this lack of being able to perform things in asynchronous manner makes “easy” thinks ridiculously difficult.
Last year we have been building a system that had to send a greeting message over Pusher channel exactly 5 seconds after user signed up. Combine it with a high number of subscribers, and Ruby’s inability to handle threads to get the idea. Similar issues will arise when you are in a need of communicating with external web services, run slow database queries etc.
Ruby does not come with tools to create pools of workers. Supervising them and handle asynchronous communication will be difficult too. Erlang/OTP does help. You can design all sorts of asynchronous systems easier in Elixir. Your developers will have to write less code. They will re-use existing patterns and libraries rather than trying to re-invent the wheel.
One could argue this is the typical area of expertise of Node.js. This is right. In my opinion, the approach Elixir takes here is way better. The code will be cleaner and easier to maintain. There is a good chance the overall performance will be improved too. This is thanks to true concurrency that the Erlang VM supports.
3. Distributed / clustered systems
It is usually better idea to scale horizontally if possible, rather than vertically. This means that instead of buying a bigger, more expensive server system, you get yourself 3, 5 or ten ones. They are semi-independent machines. The idea is that you connect them in some sort of networked fashion. Handling the incoming requests is performed in some sort of shared manner. You may want to randomly distribute requests to all nodes. This can be done in fact in Ruby with a proxy sitting in front of your “cluster”. If you want to share state between those nodes in any way, you need to rely on external to Ruby tools such as Redis.
Ruby comes with a distributed computing solution called dRB or Distributed Ruby. It’s designed in a way that exposes objects over a network. It also handles incoming requests in new threads. You probably know how this is going to turn out. Erlang and Elixir rely on message passing rather than function calls and sharing state. This makes it perfect for building systems, that are decentralized and fault tolerant.
4. High-availability systems with true zero-downtime
Elixir can help you maintain 100% up-time, while still delivering updates to the production system. This is possible thanks to Erlang’s ability to do code hot-swapping. The feature has been in use for years by telecom companies. You don’t really get interrupted calls when they need to update their software. If your systeem will rely on being available 24/7 and it’s critical you won’t miss a single request - go with Elixir.
5. Server-side push and real-time applications
I have mentioned ActionCable when discussing advantages of Rails. The fact is, that Elixir’s main web framework Phoenix had similar functionality built-in before. I haven’t seen benchmarks of ActionCable yet. I do have, however, pretty solid experience with Faye library it’s based on. I suspect it won’t scale well, especially when generating a lot of server-sent messages. We do know, that Phoenix channels do scale really nicely already.
6. Large applications
If the code base of your application is expected to be large, go with Elixir without further thinking. Elixir has a built-in tools, such as umbrella projects that allow you to split the code base into chunks. You can do similar thing with Ruby’s gems, but there is fundamental difference in philosophy behind. Systems running on Erlang’s VM can be split into “applications”. The individual processes also share no memory. They communicate via sending and receiving messages. The code is more loosely coupled by design enforced by the language itself. I found maintaining the large code bases one of the greatest challenges. Developers do appreciate a system that facilitates tools making it easier from day 1.
7. When you need smart, experienced developers
Your stack of choice can be a filter for the developers you are hiring. I am observing experienced developers migrating from Ruby to Elixir or Clojure these days. If your goal is to hire the best of the best, you may want to use the technology as a filter to get to them. Ruby’s ecosystem is flooded with junior developers. This is a result of the great demand for the developers in the technology. There is way less Elixir developers out there, but if they took their time to learn new stack, that’s a good sign of their skills.
Summary and shameless advertising ;)
If you are still not decided, I am happy to help. Just drop me a line at firstname.lastname@example.org and we can talk it over. I may be even able to provide you with a great dev team for either stacks - if you need one!
Post by Hubert Łępicki
Hubert is partner at AmberBit. Rails, Elixir and functional programming are his areas of expertise.