Відео

Cool (for real) but I think you totally missed the point about cloud service. It's not about what you're doing, but how. And instant scalability and region stuff are just a very small part of it. On your project, how are you going to update the software? how are you going to update de hardware? how are you keeping track of security issue? What would you do in case of any sort of failure? And that just few examples. All of this require multiple full time job of skilled technician. Clearly, the price will quickly be greater than cloud solution or your project will be unreliable. And on long terme, cloud with all the automated part and big scale approche are just plain cheaper. If you think "I can do AWS on my computer alone", you're simply wrong

You pronounce a lot of key terms like you've only ever read them

I like the content but your changing speech levels rapidly jumping between relatively high pitched excited explanations and just about depressed focal fry soundbites that you're stiching together so you don't have to take a breath for each sentence are off-putting.

I work for one of the cloud providers, and i often wonder about this exact thing. On my own projects i only use the cloud when its free or extremely cheap. I often fear my job will evaporate when people start migration back to self host.

well, i test in production but locally :). we restore our saas product locally with docker, even databases are restored from backups. the whole solution consiste of different services and all will work locally after you run a script. but you need at least 32gb of RAM in windows. but currently it's the best. even debug production locally is possible

you’re gonna become the next boeing engineer (iykwim)

I learnt 100 years of self hosting material thanks to you! LOVELY! thank you sire

I would rather use ansible for deployment instead of bash scripts.

Yes. Seen this kill projects that could have done well but they spent too much dev time and money trying to build to google scale plumbing in AAA grade real estate before they needed it. It's not just a server bill it's business processes and extra labor to manage it with all those knock-on expenses you didn't expect. Focus on the product working its best with lean and clean code that can be refactored easily later if you need to explode it into huge multi-lane pipelines and frameworks. If you build your app right enabling the use a simple deploy process like this vid it can be scaled and deployed anywhere in hours in the cloud if you just need more CPU, bandwidth, or storage. Transitioning to cloud devops with multi-lane CI/CD is also lower effort without crunch-burning your devs to get there. Self-hosting does not always mean onsite in your garage. You can co-lo your bare metal at a DC provider that already has baked in options for bandwidth and ISPs and huge NNIs to all the big cloud providers and can be enabled via web portal order and your bare metal quickly can have the global reach of Azure or AWS without much effort when you finally need it. At that point is when you migrate full cloud and phase out the co-lo. By then you can afford red/green CI/CD cloud devops land and fancy offices because you earned it without VC covering it. You have better odds winning at the slot machines in Las Vegas or the lottery than lightning striking and you becoming the next google overnight sensation.

Silicon valley is just a front for the Klan. Neoliberals will always show their true colors when it comes to finances.

It's very difficult to get your head round the code by looking at sections of it in a video. Any initial repo, no matter how dirty, would help people have a play. After watching a couple of videos without repos, I'm just frustrated and unlikely to return ... :(

preach!

Thanks for this great vid. This is a great intro into self-hosting. There are a lot of smaller companies in my area who are moving from clout to self-hosted solutions. This is a great way to get into those technologies and thinker a bit with them.

Cloud is a bigger scam than Enron

can you make it into a more detailed and feasible video for us? thx

can i self host with Raspberry Pi?

In general, whatever I watch or listen to. I don't like to have music in the back- ground or in the foreground. Having no music is a huge plus for me. Thank you very much.

Yes, loving Caddy.

I totally don't know all of the tools you're talking about and I'm sure that I would stumble upon setting up the very first one... But I like the possibility and the video 🤣. Im just ignorant and lack time.

Good job

Great video, i will probably set a server up in my raspberry! thanks

I like this approach because it keeps everything simple and transparent whilst being containerized so that if it ever needed to scale or go be deployed in new regions it would be pretty simple to rent some instances from whatever provider and deploy there. You're also not locked into a SaSS service when they go bust, change their prices, get acquired. You're just picking solid open source software and self hosting along with the rest of your infra. SaSS is fine to get started but once you get locked in it's designed so that you can't escape.

I was excited when we got a 150k grant for azure cloud a couple months ago... that was over when I left a small AKS dev cluster running for approx. 10 days and the bill was 900 dollars. We made our initial deployement there. Spent 8 thousand of our grant for an admittedily pretty compute heavy document processing service. Used that time to set up kubernetes on Hetzner root servers and now spend about a 20th of that while also having a more stable deployment.

Great video demonstrating how cloud actually works. What you pay for is largely convinience: securing compute resources at the speed of typing, and some pre-existing options so you have to type even less. Companies care about the first, individual employees about the second part. For a quick and dirty example there still are few flaws: Stashing local uncommitted changes seems very stupid. The deployment should fail and raise an alert. What, are you going to manually pop the stash on the updated version and hope it works correctly? The app shouldn’t build from such state to begin with! If I understood this correctly,the system isn’t fully reproducable/recoverable easily: nixOS and docker are individually recoverable, but not together! Same with the DBs. You should either leverage nix for the docker setup and handle the DB snapshotting separately, or snapshot the whole project directory so you can just set up a new nixOS instance and then jam the latest backup there. Else you first need to setup nixOS, then manually pull(more) stuff from GH, and then copy over the old DB. I guess the second point would make for a bad video and overly complex demo, I point it out mostly because since you chose nixOS, pulling the gh-creds, project-repos, build-scripts, docker services, cron-jobs etc from your private GH could have all been done in a single system-flake. And if you prefer bash-scripts for simplicity, you can inline them in nix, or again just pull them from GH. Doing this would come with some restrictions (version pinning by default) but that provides resilience (if head of main killed is what killed your old system) and is easy to sidestep (just switch to origins HEAD at the start of your script).

I'm confused how using caddy means no need to setup certificates. It looks like you hosted on your own domain, that surely needs a certificate provisioned and installed for your domain, right?

Dart is my favorite, as well as I love flutter

Not even sure how to begin understanding what's going on here

Normally I don’t hit the like button but you deserve it.

This is an awesome video! I see so many cases where teams just default to the Cloud for everything. Then, they're fighting with things like IAM and networking policies and chasing down subtle issues for days that simply would not have even existed if they had chosen to deploy to their own data center. And, don't get me started on the AWS Lambda and managed serverless idea that was going to solve all our problems - no, solved some, created others. All, while 100% locking you into the vendor. There is a place for most technology solutions. And, you did a great job of mentioning some reasons to go Cloud. The Cloud serves a real purpose. However, it's still amazing to me how so many companies and otherwise great technologists have drunk the "Cloud kool-aid" and look at you with disdain if you even mention that they should consider another option. Hats off to the vendors for doing an outstanding job of marketing.

So, instead paying to AWS (or any other cloud provider) and to a few cloud engineers, you pay for: - hardware - replacement of failure hardware - someone who buys and replaces these - someone who can restore and restart all the stuffs after a hardware failure 7/24 if needed - someone who protects the hardware from fire, criminals 7/24 - someone who installs all the os and tools shown in the video and keep them up to date - pay for internet bandwidth, pay for a secondary IP to avoid stopping business completely in case of problems with the primary IP - pay rent for the building where you keep your hardware and your system admin staff - pay for the energy bills (or spend your money on investing to green energy) - pay some data engineers (same as for cloud solution) And you still haven't done anything with your actual business/product, you still have only one server room, so good luck to scale global with that.

I don't think it's that simple, because you don't know how much is the opportunity cost of having to create your own data center, maintain it and scale it vertically when necessary. It's gonna take much more time than using cloud for sure. Do companies optimise for cost or speed? Probably somewhere in between and sprinkle some convenience on top of it all. Having 15 balls in the air as a business owner and your business growing faster than you can keep up, it's obvious you'd rather not have to build your own datacenter - that's space, hardware, software, security and admins. Does an average business owner even know when and where to start? Who to hire, what hardware to buy, how to secure the data? With cloud providers you also get their tried and tested expertise which is way more important than having a linux VM on a computer somewhere. Make no mistake, I love tinkering with my home server but cloud is a pretty awesome tool for businesses

love your insights <3

python 🥇

You go to cloud primarily because of security reasons, then for scalability. Everything else is not important including cost. There are regulations and law. Your service is hacked and personal data of your clients is stolen = bancrupcy and possibly jail. Doing security on your own costs well beyond this $100k a year.

Did nim dirty here. And making a final call on that before using lifetimes, and async/multithreading is cheating! Hands down the worst parts of that language.

What you're paying for though, is the security.

How did you breeze past the fact that home setups require something to deal with dynamic IP addresses? For me, this is adding a massive level of complexity that I cannot seem to find on any guides. Am I missing an important piece of information here??

Great.

Nice.

I’m a software engineer with more than 15 years of experience. I don’t like the mantra “cloud for everything” but you’re wrong if you comparing the time and resources required to set up some cloud stuff VS developing AND maintaining it from scratch. Like the VOD AWS, it works perfectly and let’s you save 90% of development time. And it’s just an example. And time = money

you completely missed "classical" ML where neural nets are not used - SVM, Decision Trees, KNN, etc.

Could we use a raspberry pi?

GTFO with ur europropaganda

You are an expert in DevOps. I have zero idea on this other than Vercel and Netlify

You respond with 404 when you can't map URL to the resource (the R). If you're searching for /users/1, and the user with ID 1 doesn't exist - it's a 404. If you are searching for /user or (/users), or filtered users, you're actually searching for a collection of users, and just because the collection is empty it doesn't mean that it doesn't exist. That's why the 404 is not a suitable status code. So, no - it's not a special case from an HTTP point of view. It boils down to this: if you can't map URL to a resource - 404, otherwise 200. Obviously, this doesn't apply if any other form of error arises.

The problem with this video are the “..because in my case”, “…for a simple repository with a single target” and so on and so on. There’s a reason why the cloud is so popular, it’s because it saves you time and headache. Try maintaining 10 projects in this way, deploying them around the world and having a lot of traffic and live errors on each. It depends on the service, but I would NEVER trade Netlify’s 20€ per month for this kind of DIY setup. Add in planetscale and you can do anything you want. As soon as you want to update the Linux version, something goes wrong with your hardware, your internet connection get’s f*** or any other reason you’ll regret the DIY choice. Cloud isn’t overpriced at all.

Source code? 🥹

The problem is that people over hype LLMs and think that they are Artificial General Intelligence (that they can do everything) which is not true at all. They can only do a few things, and even some of those they can't do very well, it's just impressive how far they have gotten, and people think that means that it's good at doing it.

In my opinion, a bunch of if/switch statements can be an AI. AI just means Artificial Intelligence, so it doesn't need a neural network, etc, it just has to mimic intelligence. This is what AI in games have been doing for decades, just using simple logic to simulate intelligence, by making an opponent, an NPC, etc. Though there is still a difference here, some NPCs are just scripted, while others can make decisions that weren't entirely made by the developers. So I'm not sure if an NPC that just follows you and gives you dialogue would be considered AI, but one that also interacts with the environment by for example helping you take down enemies could maybe be considered AI, and ones that roam freely are definitively AI (e.g. the NPCs in oblivion).

Curious about how ICP compares to AWS and Azure