Indie services with minimal & cheap web-presence – Small bets for backend folks on a budget

When lambda’s just won’t cut it. Just see what sticks. Get them cheap domain names.

Today’s question and related topic shall be summmed up thusly:

How can I, as a seasoned non-javascript/typescript backend dev build services requiring persistent processes on the cheap without learning an entire new stack?

Let’s Start With Why

As mentioned here, I started toying around with small projects that provide at least a marginal amount of utility.

The indie software community I take my inspiration from is nice and all, but from what I’ve seen so far, it seems to be biased towards front-end centric tools and technologies: mostly javascript/typescript for the UI with the occasional modicum of backend logic in a lambda or server-less function.

How nice, something new to learn! was my first thought as I started browsing through tutorials.

When (not) to learn new stuff

To absolutely no one’s surprise, I discovered the following annoyingly obvious fact: if it takes me a day to do something with a toolset I know nothing about, but it would take me 20 minutes using what I already know, my motivation takes a deep nosedive¹.

This is not to disparage what exists in the front-end world: I helped a friend build a simple web app to diagnose how bad your attention deficit has gotten because of your smartphone². Doing this, I learned about Vercel and Supabase and… holy crap do I recommend you use these or similar services by all means if you can! It has never been easier to ship code! I’m still flabbergasted at how fast my friend building the frontend was able to set things up, while we were discovering supabase together: it wasn’t known to us before!

The Crux

Now, to the but… The first ideas I was toying with involved some long-lived processes, and while I enjoy working with Typescript, I’m nowhere near as proficient with it as in Scala or Java.

savetoworkflowy.com is a simple twitter bot just listening for tweets, while pmon.me’s end goal is to do regular checks on the profiles of subscribed user. For both, the fastest way to build them – for me – is to have services permanently running in a docker container, written in the language I know the best.

I’m sure there are ways to achieve this with a combination of Lambdas, Firebase et al., but this brings me back to the motivation issue above: how to get to results fast using what I know now without breaking the bank?

The alternative to going back to PHP³, which was a free option available to me as I already have a hosting account that supports it, was to bite the bullet and pay for some VM and public IP on a cloud somewhere while avoiding AWS and Azure because even their cheapest offerings I could use were too high to my taste⁴. Sure, I probably sound like an old fart to the Vercel/NextJS/Lambda crowd, but the idea of having a host I can SSH into to run things as I see fit and check on logs was still a plus to me – I’ll modernise when I need it.

So hunting for cheap things I went and quickly came up with the following shopping list.

Shopping List

The table below shows the monthly costs of running my little experiments:

Why, bis

The rationale for using feather.so is that while I could host web sites on the VM or on my existing hosting solutions, these options are both too cumbersome to my taste and the monthly cost is worth it to remove that burden. Plus, it’s nice to have a shiny website up and running even when I screw up something with the VM and my services are down.

This puts me at around 21$/mo: not exactly free but, with the massive advantage that the marginal cost of adding a new project is below a doller per month, as long as the computational needs are low or non-existant⁵ – and essentially free for the first year thanks to Porkbun’s low sales price.

Of course, if something was to blow up traffic-wise, I might need to upgrade something – but that’s actually a problem I’m looking forward to⁶. Also note that, monitoring wise, the setup is very minimal and locally hosted. I intend to keep it that way as long as this is basically a hobby.

Cutting Costs Further

If you’re starting from scratch and really are on a budget, don’t care about hassle free web-hosting and are happy with gitlab’s free tier, you can cut away 14$/mo from the total and be set up for roughly 7$/mo for as many backend centric experiments as you’d want.

Should you be building things that don’t require direct exposure to the internet, such as a bot with which interactions occur via messages, you could even consider running your experiments at home on a raspberry pi or similar. Being in the process of moving, I avoided that option, and I need internet exposure anyway.

Of course, you could have your home-server write to supabase or fly.io and serve the data from a server-less solution, assuming you are comfortable with that, too.

Your Turn

I hope this was informative. Now go have some fun building stuff!