You may not need GenServers and supervision trees

I agree with you that for the most party you don’t need to use any otp stuff in servers but people should also understand some of the awesome things you can do with it. Having an distributed cache(mnesia), tracking state real time without the need to hit a database in genservers. Also depending on the structure of your code you don’t “have” to take a performance hit. Running everything through a genserver…sure its going to be quite slow. spinning up a new genserver per “user”…your performance is probably still going to be pretty good and now you can track things in memory for users. Again I agree that you, 99% of the time you do not need to use anything otp related but when you are in that 1% its really a game changer that makes elixir as great as it is.

I for example heavily use otp things(mnesia,ets,supervisors,genservers) in a saas product I am working on and it has got to the point where I do not think I could get the same performance out of the box with any other language.

Thanks for your response! Totally agree, it can be a total game changer with what you’re able to do with it and I myself still have to get acquainted with large parts of OTP. Knowing that all of this has proven reliability and performance in large system is also very comforting

Oh definitely - didn’t want to come across as saying they’re bad for performance. On the contrary, I usually think of them as improving performance (by parallelization). The fact that they’re so incredibly cheap to spin up is just sprinkles on the cake!

I think that’s “kinda” true You sure don’t have to use them for the sake of using them, but the only way to feel their awesomeness is to actually use them, otherwise you’ll just never get the feeling of needing to use them (if that makes sense).

We have a bunch of apps and only the first one, a tiny microservice actually has no GenServers. Even things that needed cronjobs / sidekiq profit from them and these are not exotic features.

I almost wrote about this not long ago but while writing it out I came to some unfortunate conclusions. I’ll just cut right to the chase and say if you’re not using Elixir for the genservers, supervisors e.t.c. then it’s probably not worth using the language.

As a language, I love Elixir. It has made some compromises but nothing major. The only real issue I have with it is that it’s dynamically typed. I’m pretty sure I linked @PragTob’s blog post in a thread the other day talking about this.

As much as i wish I could just use the language and forget about complex supervision trees e.t.c. it just doesn’t make sense. It’s not that fast, it’s not that safe, it’s not light on system resources, it’s editor tooling isn’t great, it’s not suitable for scripting, it’s not suitable for GUIs and most importantly it’s growth has peaked. The only thing it currently makes sense to use Elixir for is fault-tolerant server side apps.

If I understood @PragTob point is that if you are using something like Phoenix or Ecto you are already taking advantage of the faul-tolerant side of elixir and you dont need to try and shove otp stuff where you dont need.

As a very bad comparison, I can use Rails without any Ruby metaprogramming black magic on my code but still Rails is using it (very) heavily for me.

While I understand that, having just Phoenix and Ecto be fault tolerant buys you nothing. Any modern web framework (regardless of language) can handle bad requests without bringing down the server.

My point is that if you’re not using the built in supervision tools then there is pretty much no benefit to using Elixir but many downsides.

Not completely true, You at least get a language that runs across cores pretty nicely out of the box. Not saying other languages can’t do that but its really nice.

So what are the many downsides to it?

Previously you mentioned “is not that fast/light/safe”, which can be true about anything else as well, it basically depends on what you compare it against, nothing else. It can also be much faster/lighter/safer. So what are your points of comparison?

I personally feel very happy writing in it - I just wrote a queue system for a game in half a bunch of lines, split between genservers and ets tables…

I agree with your feeling and I would like to extend this to processes too. They say that when you are a hammer everything looks like a nail. So I see people coming to elixir becoming hammers and trying to use process for everything and getting frustrated in the process (ha!, no pun intended).

Its like getting into ruby and somehow convincing/forcing yourself to use threads everywhere when in real life you would’nt really be using them that much.

I think its an issue of communication.

Yeah but I get that with many languages now. I’m still half in the C# world and despite the hype about Erlang/Elixir multi-core usage, it’s actually not that great at it. I’ve done some benchmarks and saturating all cores caps out at about 75% CPU usage. In C# I can get almost 100% usage. Even if I could push Elixir to utilize all 100% (you can’t, it’s fundamental BEAM overhead) you still won’t see the performance come anywhere close. I’m talking orders of magnitude difference in most tasks.

What I will say is that concurrency in Elixir is way easier and safer.