Where Should You Host Your App? Hosting Providers Compared

You It's coming from Akamai, repo on it, Those are, like, the big in I don't really know a whole lot about this. This is why we want to have the AWS. But as far as I can see,

what's called wall time. I didn't even know this, but basically, CloudFormation is a description of all of the pieces of AWS. So, of course, you want, like, hosting, but you probably also want CloudFront, which is their CDN that sits in front of it. You may be also want a database.

Let's talk about pricing. How do people charge? We will go through it on each one. But generally, it's either by jobs that will run every so often. So, I first I was I looked at it and it says Based on how much you spend on AWS.

If you run a serverless function, you're only paying for the amount of time that that serverless function is running. So it spins up.

when you're working with

that really started out as, like, the Next

JS hosting provider, used option here. Now the cool thing about Vercel in my experience in terms of, like,

The continuous integration stuff is good. The preview URLs are good.

and Next JS If you wanna save costs and do more things yourself. They're not going to be necessarily

So Vercel,

kind of React in a big way. React is really,

Entire workflow

really popular.

Really became

It's the hosting provider

or Next. Js or any of these options that have an option to deploy to Vercel. I have found the experience overall to be really super nice.

Somebody who looks at these things from a user friendliness perspective, you know, I I don't wanna spend a lot of time with my hosting. I've always found Vercel's

apps that are meant to be used with Vercel, I. E, SvelteKit, Linode is now basically fully Akamai

a lot of interesting stuff. We're gonna first explain What a couple of things are. We just need to get some concepts down

FlightControl, Render, Vercel, Begin, Heroku, DigitalOcean, it's funny, not all of these services support this type of thing where you need to run some specific code every couple Of those servers to different areas around the world or

We've put some time into researching. And quite honestly, I have used So it does that all of that for you. And the pricing is based on your company of servers are they running? Are they running

of process, UI features,

Yeah. I'm surprised at how many I tried as well.

each of the different Hosting providers that are on there. By the way, the hosting providers we're gonna talk about today are

Yeah.

Fly. Io, Railway, Cloudflare, and Deno Deploy.

There's a there's a lot of variety here too in terms

I think I've used almost all of of these today. Every time I do a new project, I try to always do a new stack and a new hosting provider so that You, the listener, can benefit from my misery.

Used to be able to host straight up node apps. Right? Yep.

Vercel, back when it was, is maybe one of the most stressful times in the household. Absolute chaos. So What had happened

But they they've really changed things around and and no longer is that the case. So render There was, like, a good option for, hey. You have a node app. You wanna get it up and running, and you wanna get it up and running cheap somewhere quickly and easily With all of the the CI niceties that you would have. And in in addition to that, you can host a lot of things from, like once

for, like, an app at the time for me was just a node app, Was that oh, you can host straight up node apps here because

I have quite a bit experience, which is render at render dot Com render was an interesting one for me because the initial branding of render, especially

Zite,

and link all your GitHub and everything together. Yeah. Okay. Well, let's get into the next one, which is where I host level up tutorial. So I've I've,

Per seat, which is $20 per person, They've pivoted more to application. They have VMs, you can and I rarely get those emails, but that means that my long running process And then I got an email like they were always known for their Like articles, you know, like anytime you searched anything Nginx, So So they put you on another one. And the way that it works is that they rebuild your application One more I didn't talk about was, like, multiple environments. So it's very common now to have every time you put in a pull request,

on the back end if you want to look at when a function or how many times a function was Was run like really good insight into, like, what's going on.

Like caught on to Netlify. Netlify

looks like they're fully rebranded, You know, I I don't even know why it failed. It could have been a node version. It could have been anything. It failed. I closed it, and I have not

I can't I can't keep saying this now. Akamai is a major because is funny because Nixpacks is a open standard to deploy a kerfuffle it's It's pretty good. They do databases as well, took quite a bit in there. But, yeah, they they had a big layoff on their

Akamai. Yeah. Or Akamai. See, I wanna, like, go Japanese with it Has not changed in however long, at least has not changed in any possible ways. They have changed. They've gotten really crappy.

which you know, actually, I've gotten used to the logo. I I really No. It looks like Walmart. Don't you think it looks like the Walmart logo? Well, didn't say I loved it. I said I've gotten used to it.

as their edge play, meaning that as you need it. Like, for example, my boss monster, So I'm sitting there for 29 days a month.

In terms of its naming. Yeah. So Akamai bought the node Clicked over to Amazon and created a what's called a AWS Fargate.

really came out at a great time where, I don't know. I guess I don't wanna call the boutique options because that that kinda insinuates that they're they're small. But, flight [email protected],

Akamai is an American content delivery network cybersecurity and cloud service company. So, like, they have servers everywhere. Often when the time that you're using because of that, it was It was spiking and falling over and over and over again. So All of either a range or pinning, which means Run on this exact version of node. Yeah.

Just based on the way it's spelled, but, you know,

Akamai.

nobody's perfect. No. What about Linode? Linode, who, again, full disclosure, has sponsored this podcast before.

It was very early on. They're likely yes, this was the likely ironed out by now. But then I didn't realize that they have serverless functions. They have S3

Akama Akamai Akamai?

because Azure,

instead of Linodeau. So it looks like Linode might be going away at the dinosaur

I've entered the cold water, and I'm like, alright. This cold water's fine. I'm not like, oh, I love the cold water.

So it's it's interesting that they just Dropped the you probably have budget for that type of thing. I think that's why a lot of these companies are doing it.

And and since then, they've continued to evolve. They recently did a major rebranding,

about this stuff, but, you know,

primarily known for Deploys from a specific Git repo, whether that be GitLab or GitHub or any of that stuff, you connect your hosting provider

Linode has is often been Considered like a really good VPS virtual private server option. Leno does everything. They've been around for a long time. Very trusted.

What about okay. So next one To be Netlify. Netlify

Media Temple. Oh, yeah. Yeah. They got by GoDaddy. A lot on Media Temple. They got bought by GoDaddy, and it was like the day that that happened. I moved everything to DigitalOcean at that time because I I still love Media Temple. Media Temple was the hot hot pick for a little while in terms of where everybody was hosting.

Woah. Yeah. Yeah. So I I I feel like I'm at Linode's website somewhat frequently, so this must have just happened. But, You know what? I I'm gonna be real honest with you as I had not heard of this one until we'd made these show notes. And, I gotta say their website's Pretty cool, man, who bought Heroku?

their content in here. But, you know,

You know,

a DigitalOcean article would come up. And that's also why they bought CSS tricks. Right? But then They also bought Scotch dot io, which Oh, yeah. You go to Scotch dot io, it just redirects you to DigitalOcean now. Like so that's a kind of a huge bummer to me. It's like, I I don't know if If they just integrated all

the Scotch. Io stuff was always so good. The CSS trick stuff Is it a huge bummer? So, you know, I I I like DigitalOcean as a service and as a Deploy things. And nowadays,

you know, not a lot of hosting companies made things as easy as seamless You get in there. You get access to control that thing as if you were like a system admin for your own server. Right? And that was really cool. I remember feeling, like, super powerful, Static sites are free there. You can have workers.

Akamai.

Like, I don't know a single,

Akamai. Akamai. I don't know how people say this, but Akamai.

a technical thing, but, you know, I I tend to not, like, really, like I said, get warm fuzzies about them,

Akamai.

Linode branding because, like, Linode to me is like that's what everybody used.

But I know lots of major, major companies that will use it. Infrastructure, which is

As nice to work in as Netlify.

I've,

Anything that's like You pay based on And then I gave it access to my GitHub Repo with the syntax So it was an easy fix. I environments as well.

compatible storage, which is really cool

And, Yeah. Next up is DigitalOcean,

How they build things.

You put a website on Vercel, and it's fast. And that's because they have lots of sauce

They have a CDN as well. So if you have, they got some special sauce Amazon and all these big providers have lots of space, and you're sort of just renting I'll deal with that in like a minute or something like that. But then I also like 3 minutes after that email. Again, I hadn't touched this thing in 3 weeks.

which will build it quickly. They have CDNs which will deliver To your user, they have edge functions that put it all over the world.

move your app API. They don't have a full Node. Js API, So, like, you got to fix that before we move you over because we can't build the new version of it.

So I'm just gonna say it to you. So let well, let me explain. I haven't even told you what happened yet. So I know what the problem is. Okay. So big service providers with Play. You know? So it's kind of interesting. It's officially They were trying to move me.

on Node 18, Doing it. That's one of my gripes with Demos. Render right now. Little, own sauce to deploy it. But package. JSON And package. Json has this thing called engines, which is you declare which version of node you want it to run on. And I had I would run several sites off of 1 droplet, like Pons of sites off of 1 droplet and it was super cost effective. Yeah. Yeah, it's great. I did try the app platform as well. Actually, they even they sponsored my sticker drop as well. So Full disclosure. They paid me crashed

I have my app running

and that that's zero downtime for your application. It works great.

3 minutes after that email, I get Some others to run their stuff, Behind it, which is pretty interesting. And it's all based on an open

So I got an email from them and said, hey, we need to move it, but your build is failing.

From the message, hey, we need to

Amazon has a tool to manage its own services, and then it will bill you based on how much you use that tool. So the pricing for flight control is

uses AMS The actual CPU that you use and for some crazy They came up with this sort of, like, open standard called build packs. And then DigitalOcean all of that. But if we have a show from 8 years ago

solutions, and FlightControl trying to visit And then you also will generally pay for bandwidth input output.

AMS

It's basically tooling used to orchestrate the different AWS

To make a versatile, like, experience that sits on top of us, at least as far as what I can see it. So it's really hilarious that the other day, So I really like Arc. It's if I need like a quick I just want to run this function and I need a URL to hit it.

AMS, then you pay a percentage of that Towards flight control to build, which is kind of cool because you're not paying flight control for hosting, you're using it yourself. You're still paying All the bandwidth costs yourself for AWS. You're just using this as a nice way to interface with Cloudflare Workers and Deno Deploy do this. So if you

with CloudFront.

you either write your own CICD stuff with GitHub actions or you just hey. Hey. Whenever I push to the main branch, automatically

a lot of these hosting providers provide at least

your, Google Cloud server or an Azure server or something like that.

but you were having to manually

The information they can get, whether that be media or anything like that, has less distance to travel. You can put a CDN in front of your entire application,

it's it's Mhmm. Yeah. And and many times too When you're maybe you're provisioning a a server or something like that. Sometimes you can even choose you're having to deploy to this.

as well. There's a lot of different providers that do both CDNs and hosting, like Cloudflare,

To the repo, and,

just a a straight up server based website or any type of static assets or anything like that. In addition, there's also CI, continuous integration, or CD continuous deployment tools back in the day. You know, if we were deploying a website,

But CDN, content delivery network, basically is a way to store static assets and serve them up from

an area like The edge closer to the user. Right? So you're having you're distributing

It's something at this day and age. The CDN feels like it's something you gotta have if you are running a a normal style,

ways that you can automatically Vercel is the,

CPU.

SSH into that server. You're having to do your Git commands or, worse, you're dragging and dropping via FTP,

you want automatic deployments whenever you commit to the main? Awesome. Sounds good. Yeah. We also have Cron Jobs.

And and there's a lot that can go in there as well, but a lot of these Systems that, you know, Netlify was probably one of the first. So I know Heroku did a lot of stuff like this. Mhmm. But Netlify was one of the big ones that Really made it simple. And then a lot of folks took that same style and said, oh, wait,

And so there's there's options there. It typically, like with these things,

the c's here, CDN, CICD,

assets across all these servers. So that way when somebody loads them up,

These are the ones where you're you're honestly having to do a lot of stuff yourself. Now there's some Like we did, earlier on, cron jobs.

But, again, these these tools are going to be a little bit more of tools that you can use.

one we're talking about today, except no. Maybe not all of them. Not not digital. Not Linode. Most of these providers are running on some of the big boys, and the big boys are Amazon, AWS,

Rackspace and Vultr.

more boutique options, which honestly, the boutique options are built on top of these. Yeah, I think. Yeah. I think

Google Cloud Platform. And then there's a bunch of littler ones, not littler ones. They're major, but IBM, Alibaba, Oracle, Salesforce, A 1,000 gigabyte hours of execution, or others.

some of the the big boys, 1st and foremost. We're not gonna be talking about a lot of these in-depth because, you know, a lot of these providers,

caveats to that.

visiting that thing once a week. You know, like, it would be nice if they could spin down like Heroku did. Totally. Yeah.

Spins up the exact same server that I'm paying for in production

if you had a service that was not being used, It would spin down. It would turn off and you would the free dinos pretty aggressively did that as well, where which is kind of frustrating, and

utilities. Every single pull request I have on render

and runs It's interesting now that to build or same with Linode. You can pop their URL and the library will just work super well.

And it it seems like it's just like a skeleton crew sort of running it right now. It's funny that sometimes companies get bought

like like, did you not just spend 1,000,000,000 of how much did Heroku get bought for? Yeah. Seriously. And and you're just gonna let not only, like, dust cover it, but you're gonna let the the rust $212,000,000.

and it was down for, like, a month, like, because they had a security incident.

I'm gonna be polishing it every day. I'm gonna be washing it. You know, I'm gonna I'm gonna be taking care of it. I'm not gonna let it fall into disrepair like, this has. And so, Heroku is probably only sustained by the customers that they had in the past. I I I don't know who's launching on Heroku today. If if not there, I recommend it. Heroku fan number 1, I guess, reach out to us. But, I would love to know if anybody is launching on Heroku today when there's all these other better outlets out there. Heroku was the first you know, primarily, at least initially known for their droplets, Which are essentially just a VPS

the The free ones, is there even free download anymore? Are are totally gone. They got rid of that. They've had major downtime. People were running that for all kinds of they had, like, a git integration,

Is that, like, if I bought something for $212,000,000,

You can't like the, their free static site hosting. At least that's why people initially, kind of rode that wave and and became a really highly

Yeah.

them as a a company necessarily given the CSS trick stuff,

used it since.

that happened. But, you know,

they're definitely, like, a good option and a reliable option to host on. And if you run your own for a long time, I personally hosted level up tutorials on Heroku because before Meteor had their own hosting service. That was like a really good place to host a meteor app. So I I had a lot of stuff on Heroku.

doing that kind of thing depending on what type of services. Also Yeah. You think, like, Kubernetes, Clusters and stuff like that. So, I mean, they're they're a legitimate option for,

big time, small time, anytime apps. Yeah. They do

So I you know, DigitalOcean, they're very reliable. They they have a lot of Great things. I I don't have necessarily warm and fuzzy feelings about

Server on there. You're your own VPS. You know, you you can, like, really save a lot of money,

went into my or or they know they take the built version of it and they just cut it over to another one. And then they just switch the domain name Because it was failing, but they just moved it copy pasted onto node 20.

as somebody who's who's I remember when Heroku first came out, their website was

the all of the issues for me. I take back my judgment of you, Wes. I think you're fine. Okay. Good. But I was it's It's so stressful when that type of thing happens, especially with, like, memory things because you're like, what? Yeah. It's not just like, oh, there's a syntax error Or like, this is what happened. You can look at the logs. It's just like you're out of memory. Boom. It fell over. But But you get the whole you get the deploy logs, you can have multiple although serverless can also be a virtual machine. But the idea is that you have a long running server. You start it. It runs. It handles all the requests that are coming in. And then a lot of these providers will also allow you to

They took

is So it's

and x minutes, hours, days, whatever.

But also Super quick ones. I like ARC because it has routing bait built right in as well.

node greater Or equal to 18 point whatever point whatever. Right. And I said, like, that's a minimum I want it to run on Node 18. But the day before, Node 20 had been released.

It was failing. So I fixed that. But that doesn't explain for AWS, meaning that they have sort of built a

So it was trying to rebuild it on Node 20 and failing because, like, I hadn't tested it. I hadn't upgraded anything like it's not no 20s fault. It's just that You should test things on new versions of software before you deploy. But because I had greater or equal to and not $50 or $100 on

copy pasted, Yeah. So it's it's an open center. It's pretty cool. Now I never heard of that before. It's interesting that it's open sourced. So Heroku has build packs, which is, Okay. I have a Rails app, and I want to deploy it to Heroku.

don't do that. Like, if you're going to move me, maybe copy all the specs exactly how it was running.

but they didn't

my built image it deployed the syntax website. And Just finishing dinner, trying to get the kids in in in in the bath. So the worst time I know. To have anything stressful happen. If anybody has has kids, y'all know that bath time

all of a sudden was failing.

just took the I just pinned the version to note 18, redeployed it, and it fixed Akamai.

I would say mostly my fault, Fell over somehow, right? So I was like, oh, shoot, like, what's going on here? Welcome to Syntax, the most hosted website in the world. I don't even know what that means, but today, we've got A show for you where we are going to compare

And what I did figure out, what I think happened was It's not a hosting provider itself.

and server full CICD,

You're listening to Syntax, the podcast with the tastiest web development treats out there. Strap yourself in and get ready. Here is Scott Talinski and Wes Bos.

CDNs, cron jobs, auto scaling versus not, CPU time, wall time, all of that good stuff.

all of the hosting providers for where can you put your website. We're gonna talk about

Serverless edge AWS Lambda functions. They do edge functions, which are just Cloudflare workers, not just, but like If you deploy middleware in Next. Js, for for what you get on the base plan. For $20, you get you get a lot for that.

an instance of that so you can actually see the output I was at the time,

What happens when you get lots of traffic? Does it auto scale? Can it auto scale? It will spin up as many as it needs to handle all that and then it will spin down. It does stay warm, meaning that the function stays

Of deploy click on the website. Deploy previews. So most of these services now do that as well.

Do you need to do that yourself? Do you have to decide that? Mhmm.

Yep.

That serverless function might To dormant and you're not paying for it. But Maybe you're saving some stuff stuff to a database, and then you send the response back.

deploy multiples for as long as until I can merge that pull request inside of all of those, there's kind of a couple main ideas. 1st of all, we have what type sometimes it runs 1, Sometimes it runs 6. Depends on how many people are my whole instances was scaled up to 2 instances, and they were So you just have one that's working. Yeah. Oh, they were Toggling back and forth. They were just both failing

serverless function will never run. Or if I have like a PDF generation where once once once a month, People need to go download a PDF of their bill from their cell phone provider the the saving grace was I actually didn't have any downtime however much. And Vercel actually gives you

provider time from them. So that's serverless. Then we have server full, which is the opposite idea where

as soon as it's needed, it's not going to be slow. It's not going to have to take up space on a rack. It will just spin up. It uses the idea that Netlify does it. Pretty much every hosting provider now says, Okay, you pay for usage or like your base plan gives you Turbo Pack, and you could scale it up really easily. Like, I remember I was getting tons of traffic. All I did was

the thing all at once. So if I send a big email out, The thing is going to get a spike in traffic, but I could just spin up more server full slapped it into a node 20 environment, and started again. They didn't rebuild it

and you run a server. That thing runs until you kill it. So

Like when you run a node process on your computer, you type node index.js what what do you call Akamai? Cloud?

that Nobody touches

and it was, like, the first time you're learning Node. Js. Get something up on Heroku in no time. That's like the whole whole premise of everything that we did. Yeah. But

They've gotten really crap. You're right. Yeah. Has not changed positively.

They got a lot of people roped in with their free dinos, they called them at the time, or their dinos. I I I still think they would still probably call them dinos, but Yeah. They're free servers, essentially, that you could quickly deploy an app on. I remember going to meetups, And and sometimes it's simple. Sometimes it's automatic, and sometimes A simple way to spend a lot of money if you just say, here's my credit card and scale up till infinity.

10 seconds you might want to pull an API and Check for new instances of something so you can

For example, you might have every morning at 4 a. M, you might run a database backup or every send somebody on their way. So those are sort of the 3 big different places. Yeah. And then Other topics, we we started getting in the

So crown jobs are good. Scaling, I talked about that earlier.

Cron Jobs are

So this one, it creates a cloud formation, which is

service.

And I was like, what? That's that's a lot. But then I read closer, and that's not It's based on how much you use on AMS,

bucket. Test 3 bucket, you know.

You might also want a bunch of different of these, like Route 53 And if that code needs to run a lot of times,

fame.

where exactly this is hosted. Right? Sometimes they'll give you the app. Yeah. An AWS server verse and, Heroku

10,

created from Brandon Bayer of Blitz.js AWS

Let's Get into, like, the first of these boutique options.

Have full control, but I did not want to be a system admin. So I've kind of moved to,

To attempt to deploy a personal site or something, it was like a Gatsby site.

cheap way to host a lot of stuff that allowed you to essentially Dark. It was cool. It was purple and, like, just really cool everything about it. And then you get into the actual service, which was like, oh, you could do these quick add ons. Deploying is as easy as connecting your GitHub repo and adding the Heroku deploy or whatever it was.

being in control of my own server, being in control of my own NGINX and everything like that and And really, like, dialing in that control. It was such a a nice honestly, I I I only use the app platform

which is it's a lot And it was it was running on render fine for 3 weeks. I hadn't done a single commit to it in 3 weeks.

But then you got 16 members you wanna add in that can get it gets pretty pricey pretty quickly, but I think they've Sort of nailed that because if you do have 6 team members, You can just use the AWS libraries.

Yeah. Totally.

Next one, begin. We've had Brian Larew on the show. Begin

It was $5 a month, a framework for deploying to AWS

but I wasn't I again, I also had some issues with the app platform as well.

Drag the slider, and I had to power it down, drag the slider, go I used to host my boss monster on DigitalOcean.

to be able to get a server that runs on an SSD

money, CI tool for deploying to AWS.

it was, Heroku, which Heroku

But for the most part, this thing can do

What had happened I'm gonna say it is your fault Because this is not a problem that I had.

Yeah. So it's a long running server full I get a another email from a runner said your build, Your server But I like to think of them as like a remote control

anything you need

You can host Postgres databases and private services in general there, but I hosted our Redis instance there, and it was super easy. A private Redis instance just Spun went up really nice and quickly, connected to it. No problem. And overall, had that exact thing going on before with this server is just dying for some reason while you are trying to figure it out, It's definitely a, a pro to have multiple instances running in that case when you're, when you accidentally have a, a bug make it to production in that regard. Exactly.

increase the number of CPUs, the amount of memory. So I also host My app on render as well, Linode, Netlify,

and I've been fairly happy with it. I did have a little bit of

For apps, I choose to use render as an option when I have a straight up node app that I wanna host somewhere as a modern solution there instead of going to something like Heroku.

it has been a a good host for me. The pricing has been fairly reasonable even though they did upgrade their their pricing recently.

which is always a little bit stressful. Mhmm. And it was I think it was mostly my fault.

And that like, Dev that would use Akamai

serverless functions, which, by the way, serverless functions are The idea that your code lives inside of a function. And when that code needs to run, it goes from its dormant state to running

Like, if you go to one of the pages on syntax,

then many instances of that function will run. So I always give the example of,

a serverless function will spin up, render that page. And if we get a 1000000 people try to visit that Single page at once.

very quickly.

That's really cool. They also Edge functions have a limited

And When we did the show on SST, first before we get into a a worker was a So I was like, Okay, They obviously have

The actual tech behind Vercel, they do serverless, which Is,

It's fast. I found the pricing to be fine for me.

it's one of those services that continually sees improvement. And and as we record this right now so this Could be come out of date in some way. There's like a Vercell event going on at this very 2nd where they're announcing who knows how many things or something going on. So,

Cloudflare Workers, which is the Amazon managed They're going full like corporate.

But for the most part, trigger Salesforce.

Yeah. There could there could be more here that we're even missing by the time you're listening to this.

They've sort of pivoted now to More serverless. Specifically, wasn't it? Yeah. Yeah. It was it was entirely serverless.

full stack apps,

So they are both a way to deploy it, but they're sort of pushing You're only paying for

full stack apps with web components

Heroku used to be like

Oh, you you have it in here in the notes. That's fun.

Arc is Arc or Cloudflare Workers are usually the 2 every 2 minutes, but they were not failing at the same time, and my restart time is, like, 3 seconds. So

And I I don't really have anything negative to say about Linode. Although, you know what? I just went to linode.com.

Next step is not managing my own server and doing a lot of these other ones. And DigitalOcean kind of also realized that not everybody wants that. They they were seeing the rise of services like Netlify and Vercel and and those types of things. So They have their own app platform, which,

Salesforce. Okay. So Heroku used to be

that I have 1. I have a recipe worker. Someone hits it with a URL for a chicken. So I take in

to me, I'm not paying for that time. So if the recipes API takes 6 seconds

make that work.

Loaded and ready for the next request. Prepped. Yeah. But you you only pay for wall time. It was just clock time.

to to do my process, I'm not paying for that. But as soon as it comes back, I'm Jason Parsing.

and then I send a fetch request They have good analytics

I am looping. I'm mapping. That's CPU. I'm actually using it. So you only pay for that is we call those like virtual machines, logged in and then I Is you only pay for the amount of time that the computer is doing work

And not time that you are waiting. So specifically, And so I jumped in and I was like, what? It says it's out of memory. I was looking at the logs and it was just basically, like, standard called Arc.

And while I'm sitting there waiting for the all recipes API to come back, By by the box like you would shop for a computer. I want a computer with I want a server with this many CPUs in it. I need this much memory.

way they can but that's how literally every render does this. They switch to that.

The URL. I parsed the word chicken out of the URL, every single to

are

Use what's called CPU time. This is pretty interesting.

to the all recipes API.

or a virtual private server where it, like, provisions you your own server, sends you your log or your, your login credentials.

which DigitalOcean,

taking 3 minutes. It spiked the memory all the way to what I was paying for, And then it ran out of memory and fell over, and it restarted itself. And it ran out of memory and fell over. And it kept doing that every 3 minutes.

You're serving up a CSS file or something like that.

You will you'll pay for that. Yeah. Sick. Well, let's get into the actual providers here, and we're gonna start off with,

And then the third way is you pay by Why my app was And you're paying for 7 servers even though 1 person is literally 7 servers, with their $5 SSDs, and a lot of people moved over to that.

set of tools. It is a service that you Log in with your AWS. So right before the show, I signed up. I

That makes it easier.

It is a Like most things now, they initially were $5 SSD server, which is unfathomable I

Everybody was like, use amplify. Use amplify.

And, and then it will spin down. So you got 6 pull requests in. You're running I think they had moved it, what this is, Do the exact environment because my package, Jason, said it's okay. It can run on anything bigger than 18. So Brandon came out and said, hey. Like, I'm working on a startup called FlightControl right now. And it's pretty interesting in that They have different stacks. So if you spend I'll deal with that.

but I don't know a ton about it compared to many of these other ones. Yeah. This this one popped up in my feed because

Next step is is, JavaScript sweetheart, Vercel.

So, yeah, big fan of render still, though. But, yeah, especially, like I said, for node apps, that's, like, usually the 1st place I'm I'm reaching only because it's Quick and easy to get it up and running there. I got stuff there already.

So it's pretty cool. It's using this thing called Nixpacks, which have a different API than than Node. Js. They do continuous integration. They do all your builds, all your previews.

and many other and they are often used as just in between functions as well. Or you can intercept some data and do something, That's where I ran my 1st ever server. And then DigitalOcean came around And then,

And it's built by another one on this list, which is Railway. Io.

Applications, you can redeploy it when you have a git commit. It will trigger

That's why they're open source, right? Yeah, the beauty of open source.

a rebuild and deploy it for you so You get that whole experience, but on your own except that have on demand servers, meaning that hosting images on Amazon S3, you could just use that library and and Pop DigitalOcean in there because we're doing something with those services, which is fine, Because what happens is they put you on a VM. It's sitting on a server somewhere. They need to power that server down.

AWS AWS, Azure, and Google taking up a majority of that pie.

Which

any type of language to off of the node 18,

AWS.

Providers say, okay, we're also going to to do build packs as well. So it's kind of funny because they use another hosting provider's We'll talk about the downsides to that as well. And then finally, we have Edge, which

to where the user is running. So if I run it, it might run-in Toronto. If Scott runs it, it might run-in Denver. If you, The Australians run it. It's going to run-in they're deploying those things to If you hit a URL, sometimes the 1st time you hit the URL, it would take 5, 10 seconds. That's because the server was starting up again. It was put to sleep. It wasn't Actually being used. The VM was turned off and they needed to turn that VM back on. But that's also good if you have things that don't necessarily need to be And The the function closer

like a Yeah, Melbourne.

it runs and they bought the note to sort of go into their edge Let's say you have Fetching a database. You get some data back. You're rendering it to HTML.

Edge is very similar to serverless,

It was really interesting because the Syntax website is running on my AWS account. It's not their AWS account. It's not running on their servers. It's running on mine.

This is probably not fair to Vercel because they're a lot more than this, but they've built a Vercel like experience.