CODEWITHSUNDEEP
node.jssocketserlangerlang-otp
Sergei Basharov
Sergei Basharov

Reputation: 53830

Sockets server made with Erlang vs others

I am learning Erlang and trying to understand how its sockets work as it is meant to be one of the strongest parts of the language and OTP.

I have experience with NodeJS, and wonder, how the applications made with NodeJS and Erlang differ in regards on how multiple sockets connections are managed.

As I understand, while JavaScript is single-threaded, V8 manages all the multiple simultaneous connections for it, though Erlang can manage multiple connections itself.

So, I wonder, if Erlang has excellent support for managing multiple connections at a time, how is it different from other technologies for a programmer? I mean, when I write an app for NodeJS, it can have as many connections open and well-managed as if I wrote code in Erlang, isn't it?

Please share your thoughts, links to some articles about the specialties of Erlang in this context are welcome too.

Upvotes: 0

Views: 938

Answers (1)

vfsoraki
vfsoraki

Reputation: 2307

I am by no means an expert in Erlang, but I think I know Erlang and NodeJs on the same level.

The things you say, are all correct. Bot can handle multiple connections very efficiently, well well-managed you say.

But the thing is, the problems are not only handling multiple concurrent connections. The problems Erlang tries to solve very good, are fail safety, and distribution. I don't think NodeJs will be as good at it, as it is now.

Don't take it wrong, I'm not saying no one can code a distributed app in NodeJs, but considering the tools Erlang gives you, it maybe is a better choice.

For fail safety, as an example, Erlang let's you link your processes, so when one fails, other also fails or gets notified. That is not very practical by itself, but when you look at it alongside supervisors and shared-nothing processes, it is a great tool.

For distribution, Erlang let's you link nodes together. Linked nodes can talk together as if they were on the same machine, and they can spawn processes on other side too. Consider this, with the ability to start a failed app from a failed node on another node that is healthy. Gives you a great uptime.

And not to mention that these tools have years of experience behind them.

Just try to solve these issues on another ecosystem. I say ecosystem, because Erlang as a language is not complete, but the tools and frameworks (mostly OTP) have to be considered too. Then you can also say that Erlang really shines in this areas.

But Erlang also is not very good when it comes to linear processing, number crunching, image/sound processing, etc. That would be better implemented in another system.

I think, in this areas, the big difference between NodeJs and Erlang is their runtime model. NodeJs has one process, one thread that is working async on io-related tasks. Of course, you can run multiple processes, but that is the basic thing. On the other hand, Erlang has a VM called BEAM. Erlang uses special processes inside this VM, very light processes. BEAM schedules them itself, because they are not OS processes. This gives BEAM the advantage to have hundreds of thousands processes at the same time, each doing a task, be it io or anything else.

You see the difference now, I think. Erlang is more battle-tested, more better when fail safety or distribution is a must. NodeJs maybe better when you need faster development, and deployment.

Upvotes: 2

Related Questions