We are in the process of inheriting a project whose development our client outsourced to another company, and the sheer excess storage they have for the user files one of the APIs handles confused us at first. We asked them about it, and their reply was accompanied by a shrug. It's not much, until your factor in how little data users have uploaded since launch. Idk why they are already paying for that much. May as well give us and therefore me that money.
More pain with cloud. Expensive engineers, subscription based, less secure. It's only less pain for management / ceos. "It's amazon/Microsoft's fault" Instead of "Our fault" As tech competence grows it'll become a shallower market.
Mate, I launched an EC2 instance the other day to test something for 3 days and even though the instance price was clearly around $300/mo (which was supposed to be $30 for the 3 days the instance was on) I quickly accrued a bill of $152 😂😂... against my initial prediction of a few bucks lol. Turns out I had maxed out the io2 block volume attached to the instance and was being charged an insane amount just for that alone lol
As I'm currently developing Azure-based IT infrastructure for our new 'Ai and Analytics' team, seeing a video that describes not only the pain, but the dependency really puts me at ease, knowing that it's not just me.
Part a really really large org that is multi-cloud. My team is responsible for all public cloud architecture and strategy. This is 110% it, butttt at certain scales you MUST have an exit plan (due to compliance reasons) to exit vendors. Previously we've done this via kubernetes clusters and various management services for all the core functionality. Now we have corporate policies to handle our exit :(
I'm sorry but WHY. WHY bother to learn all that proprietary nonsense that's all going to become obsolete in 6 years. I'm sure you probably need way more data than my little company will ever use, but I've tried using this service and immediately realized that it was geared toward massive companies only.
@@lashlarue7924 It's all use case. If you're making a little app to sell to people you can probably see the benefit of using a cloud provider to host it. If your business is doing something else then the cloud isn't probably the right move
The problem to solve isn't setting it up, it's hosting it in a reliable location with good AC, reliable (fault-tolerant) sparkie-sparks and of course giving it access to an Internet backbone. This is what you're really paying for with cloud tech.
@@karmatrainingexactly this. But also more like infrastructure costs like rent, security, keep it running 24 x 7. For AWS to run Amazon must have its own engineering team which keeps it running 24 x 7. Operational costs. Cloud is a hectic affair.
As a Cloud Support Engineer, the billing part is spot on. We discouraged to advise people on how much something will cost and just point them to the fancy calculator which itself is like “yeah dude get an estimate” and then people get lost or confused in documentations that are written like some old riddles.
There's another consideration why cloud is popular: cloud costs can be accounted as operational expenditure (opex). This simplifies accounting and reduces a company's income tax. Although it leads to lower profit in the long run, it can juice short term profit and keep investors happy for the next earnings call.
@@xmorse accounting rules require you to handle one off purchases, like servers, differently. These capital expenses (capex) aren't tax deductible immediately, but over several years (depreciation)
@@manishm9478hmmm, so what I'm hearing is that the company doesn't plan on being around for 5 years? AWS is paid month to month. I calculated that for cloud computing of 1 Ryzen 7 1700 (I know, not a server, just stick with it) of 8 cores, 16 threads at 3.2ghz, it would take exactly 1 month of 100% uptime to just pay for the parts yourself. Doing the scaling of nonconsumer grade servers, I can't imagine that aws would ever be cheaper than DIY, even with paying your architects $1m/yr each. Especially after 5 years. Plus tax breaks. Oh, and being subject to Amazon playing the monopoly game (or cartel game) and randomly raising prices. Yeah, fuck that. Own your shit.
There is also a business strategy that jump started and now locks you into the cloud, cap ex vs op ex. Multiple companies especially those with "rotating" CEOs look to increase company value by limiting the capital expenditure during their tenure.
I have a radical idea but the CEO won't like it... Maybe instead of a $120,000,000 salary, they just buy a server and set the next CEO up for success. They'll still have a few lifetimes of money leftover.
@@andrewcook_ That is my point salary is also a opEX not a capEX. Buying something takes a longterm commitment but shortterm cash both of whcih any good business will avoid, if at all possible. You can fire a CEO when they fuck up, but you are left with that crappy server that everything runs on and needs a team to manage forever.
I'm addicted to the cloud because i love Infra as Code. I used to work as a sysadmin and almost had heart failure several times due to dying hardware. Never again! Totally worth the bill.
Just within AWS offerings, we had a major tech project to move to cheaper solutions and tied it to our annual bonus. We did the work and saved so much money in three months that they paid our bonus immediately and refreshed it with new goals and another bonus.
The problem is once you learned how to query and setup these clouds and your app is running, you DONT want to change a running system. I would rather tolerate to pay more, instead to have a 50% chance to break everything. This might be true for personal small projects, but also to mid-size companies with 5-10 devs.
Congratulations, you just described why 90% of modern programmers are just hacks. No clue why anything works and to afraid to change a thing. Because Stack Overflow has no answers.
As someone who holds a Masters degree in computer science and has been in the business for over 14 years I call tell you “running your own because it’s cheap” is not a good solution. Cloud is expensive? Yes, but it takes a huge burden out of your team. I have worked in all sorts of projects, from bare metal, hybrid and all cloud, and there are advantages and disadvantages to all of this, but running your stuff has huge drawbacks and embedded costs that most people don’t think about. What about backup, data encryption, disk and hardware replacement and hardware monitoring? What about managing disaster recovery plans? You need a whole crew just for that, and if you want competent people, you might as well spend a lot of money that could just go to cloud. If you are a very small company it’s probably a good idea to run stuff in house in the beginning, but in the long run you want to move to the cloud. It’s just more efficient.
I would have said the opposite... If you're a small company you can't afford a whole team to run your servers. But for a large company the costs of a cloud provider might be larger than the cost of manning your own internal team. I guess it also heavily depends on what exactly your business is.
@@ToadalChaos this is literally it. They lull you in with the free tier and all its bells and whistles. Then when (if) you grow you’re locked in paying 18 million dollars. Its a part of their game and you’re meant to lose
One unspoken effect of cloud is career prosperity. With legacy tech, it’s very difficult to switch career, whereas with cloud you get more chances to enter the field because it’s so fast moving. I myself was stuck in a mainframe job and cloud gave me the opportunity to have a better career, and I see lot of folks doing the same.
I work at a rather large games company. We rent rack space at three Datacenters with ca. 6000 VMs running on idk how many HVs. Moving to the cloud now would be extremely expensive because our games are built yeah in a way that would be very expensive in the cloud. But we have a disaster recovery plan which involves spinning up a replica of our own infrastructure at AWS anytime we want. It would be supa mega expensive and would not be viable for long but better than like not having our games online :D Other publishers in our Group use fully managed AWS and pay small sums even compared to self-hosting. And they partly have more players than we do. Thats why my company is trying to figure out how to build new games in a more cloud-friendly way. Wither way the games themselves need basically no maintenance and are at very high uptimes. Only problem are services like hadoop, bi and our wallet database hahaah
0:40 To nitpick a bit, S3 and EC2 surprisingly weren't the first AWS services to be launched. Surprisingly, the first one was actually SQS which was originally launched in 2004, though it only came out of beta in 2006.
This video was AMAZING! I enjoyed everithing of it. I have thought a lot about this. I have my infrastructure on ubuntu and docker containers in rented VPS'. It's as simple as that!
I'm currently using Contaboo, which gives you the root server and you do whatever you want with it. 4 vCPU cores, 6 GB of RAM and 400 GB SSD for 4.5 usd / month. Pretty good deal if you ask me.
Liked the business lesson at the end of the video. Solutions that relieve pain but that costs money. So companies wouldn't exist without pain. We are living in a painfull world
Depends if that how you'd like to look at it. I see a beautiful world with open source projects happening that are countering malicious practises. And that's also part of the end of the video. 😁
This video is missing a huge point: Cloud gives you some super exclusive tech at a fraction of the price that it would take to stand it up yourself. If you are building a fault-tolerant world-wide accessible app with local regional edge availability (say a backend for a phone app that works globally) it would take you millions to match AWS performance and reliability. Also, their DB master-replica technology is pretty amazing that, again, would take some very expensive hardware+licenses to implement locally. So, for applications like that, it would take SIGNIFICANT economies of scale before self-hosting becomes cheaper.
I don't think he's saying cloud is inherently bad. But his videos do tend to oversimplify as he tries to make them quick. But saying "Bare metal - pain = AWS" is a lot more straightforward than showing a graph explaining how this changes depending on a company's size and the amount they already use cloud solutions.
You can pay for hosted colo, edge and run your own cdn caching layer all over the globe for $200/ru/month which way cheaper Amazon ec2 large compute at $800/month. I did the math for startups, you will be sacrificing upfront dev time on setting home grown or open source api, security, backends, broker services etc. we are talking about compute cost on leasing a server vs leasing VM, or are we talking about leasing a data center?
As a developer who also works in Infrastructure (IaaC), most people don't realize what does these 3 big public cloud providers are data security, security compliance and support. I myself working mostly with AWS cloud you will appreciate how convenience having a live chat support.
I'll go the hardest way. Not because I'm addicted to pain or I don't like clouds but I want to understand how it works from the start. Practice is the best way to learn.
@@tylerlaprade642 unfortunately, I don't have a degree in this sphere. I used to learn electronics 4 years ago. We were taught how memory gates are constructed but nothing more. Now I work and program in C language but still don't know low level enough to code well. Even though I don't know ASM well and you tell me to build my own microchip. One job at a time, my friend. 😁
It's all about time in the end. You can spend it on your actual product, or on all this infrastructure hassle. Our time is limited and non renewable resource.
it seems to be similar to microservices addiction...people start designing their applications in Microservices & start putting the infrastructure cost & then get into some cloud like AWS or Azure or GCP & then declare that they have created a marvelous design for sustainable product but that product never kicks-off properly to recover the cost of that design even ...and I feel that is another truth of cost over shooting for max projects.
We were forced to move one project from "dedicated" VPSs to a cloud platform, because the entire corpo was doing it. After spending months on that migration, I asked them if the cloud is free/included with the infrastructure. They did a bit of math (they read my math) and decided to switch back to dedi. The project evolved during that time, so it was another few months of migration. I will make it my mandate to talk companies out of the cloud, until it actually becomes cheaper and more open than hosting your own.
All cloud means is "someone else's computer". So if you're renting a computer cheaper than it would be to buy your own, then your cloud provider is a moron or is waiting for you to get hooked so they can jack up the prices.
I can definitely relate to this as a programmer. It's amazing how companies prioritize spending more on "cloud" instead of investing in optimizing their bills or hiring more engineers. The real question is, how much pain are we willing to tolerate? Every programmer/developer knows the struggle.
This is good advice and I agree. I use Nomad to schedule containers across cloud services. That said Lambda being largely free is fantastic and you can write them in a way that lets you migrate to a container solution once you scale to a point where they start to cost money.
The one thing I don't get is what the heck are people doing to complain about egress costs? If you consider something like Kick which uses AWS, then you should be kicking your behind with those egress costs because you are streaming actual video to your audience. But for 99% of businesses it's just plain text and you can transfer a heck of a lot of plaintext for cheap. So I don't really know what these guys are doing. All the resources are or should be allocated in CDNs which have different egress rates specifically made for the purpose of serving media. People are losing braincells, or are dumb to begin with.
>Nomad to schedule containers across cloud services. Do you have any data? It has seemed to me that multi-cloud falls apart once you have any sort of datastore; now your data exerts gravity in whatever system it is hosted and you naturally migrate towards that system for everything else.
@@Toramt yeah good point. I use S3 for most static files (like user image uploads and such) I also use AWS cognito for authentication (nothing comes close when it comes to price and features) Databases can be managed in Nomad. Everything deployed with Terraform. It starts getting really difficult when you scale to Facebook-scale multi region users - but for most use cases you don't need anything complex.
@@KristianRobertsen For Wordpress Admins, yes. If you ever updated a ceph cluster in production or operated a multi-region Cassandra cluster on-Prem, cloud is definitely more.
@@mackster85 depends on your skills, i personally don't care if the k8s cluster is running in the cloud, self provisionied or fully managed or onprem, rook runs everywhere, same like tekton and argocd and if i need vm's, i host them too on k8s, kubevirt
This makes me kind of glad I never got hooked on making personal cloud-based projects. One attempt at understanding AWS firewall rules was enough to scare me off
Magnificent video! Absolutely marvellous. Good take, we should always try to consider all of our options other than the cloud before picking where to deploy
I've recently started to wonder how peer-to-peer file sharing would work for cloud computing. For example with something like Netflix, when you stream a show as you download the content you'll begin to upload it again to the next viewer. The show could stayed cached a bit longer so that it can seed as much as possible. That way a streaming service would just host a UI, some torrent links, and seedboxes for when the leechers outnumber the seeders. In theory that'd save energy because the viewer's device is already on, it'd save bandwidth for the company as they biggest stress would be the seedboxes which is still significantly less than hosting everything, and it'd lower latency because you could download from a neighbour instead of a server across the country or even the continent. Torrenting obviously wouldn't work for a database since those are heavily personalized but I think whoever can make a legitimate legal and secure streaming service that uses torrenting will be a very rich person. They could pay more for exclusive licenses, undercut their competitors, and still take home a wider profit margin.
Netflix already did as you described. They have caching servers distributed to various ISPs. Works out quite well as there's only 10 or so shows everyone watches at any one time
It won't work because ISPs can fuck over the chain in quadrillion ways. Especially since a lot of them separate "external" IPs from "internal" ones, so things like IP banlists become meaningless for large enough ISPs. This kind of setup also requires extra fiddling on user's end, which majority of won't bother. "Streaming" on torrents already exists, that's what "sequential download" in torrent clients is. And guess what, it promotes hit&run behaviour and will get you locked on any relevant tracker, so no one has made billions off it yet. It's a general illusion of grandeur related to P2P for some reason. Just because a P2P network can have big numbers in throughput stats, doesn't mean this entire throughput is freely available to any given participant at any given time. In a P2P network the "competitors" are the 20% of participants who make said network valuable in the first place, so this get rich quick scheme is self-defeating in its core.
Asymmetrical network connections break this, 100 down but 5 up. Maintaining any sort of Quality Of Service for a real-time process when your 'providers' are random end users is very difficult.
@Toramt Reliability increases with more people, but the problem is getting to that scale. You probably need everyone to have gigabit fibre + terabytes of storage for this scheme to work, and we're back to the problem of crypto requiring tons of work to do what central databases do using a fraction of the resources
@@Demopans5990 I've been able to maintain a positive seed ratio using a network with 100mbps down and 30 up, only about 6TB of storage. An episode tends to be about a gigabyte, so it would be at most only a few GB at a time since the episode would auto remove eventually. Even if everyone only did a ratio of 0.5 that's still significantly less bandwidth usage on the company's end which is a lot less cost.
@@TealJosh You still own the data even when you push backups to a cloud provider (or several). The point is you should not be 100% reliant on the cloud provider and use it as your main data storage. I'd argue you're not doing 3-2-1 anyways if you're just using AWS, because your 3-2 and 1 are all based around trusting that Amazon's datacenters do all the work for you.
@@swojnowski453 True, but with banks you get the government bailing you out with tax money if the bank should fail like Silicon Valley did lol But i fundamentally agree that there are many scenarios where keeping most of your money in a bank is one of the most regretable mistakes someone can make
During an internship, I've had to learn AWS and build a fully micro-services based project (that revolved around AWS itself) with CI/CD pipeline and everything. Can confidently say I am NOT addicted to cloud computing
Haha 😂 It's just a clickbaity title :) There is a pragmatic business reason for using cloud. Business is not charity. They are not emotional about it. If spreadsheets show that it is more efficient to run your own infrastructure then businesses would do that right away 😀
@@egor.okhterov That's only one half of the truth, there is, depending on where you are, also a shortage on people capable of doing so. Furthermore, when the time comes when it would be more efficient and they would have the resources to run their own infrastructure they are already locked in. (The egress fee will prevent the switch from then on)
This is high value information from (1 whole info sector) insider. And its full of redpill humor and eastereggs. Wow just wow Jeff we appreciate you a lot! Please make sure, that no matter what, we can still watch your content, even if something were to happen to this platform !
@@boumajohn I'm not proposing a solution for owning all your data, just a midterm between investing a lot of money in self hosting and not being raped by egress costs. There is a huge gap in between, and I agree that renting isn't owning.
@@jesteriruka4215It's because of the perceived benefits gained (security, maintenance, and convienience) when a company starts spending small. At some point when the cloud spending exceeds the staff salaries, that is when reconsidering cloud becomes forgotten due to how long for many companies to cross that line. I hate to say this but usually, the people that prefers cloud to on prem or a much more plain providers like OVH or Hetzner assumes on the efficiency of wasted resources when not used. In my opinion, that is valid but what these people often forgot to detail is how their software is often modelled after the cloud specific service that made it cost efficient while also expensive should they grow with it because it is designed like that. Traditional infrastructure providers fits well with most use cases but many simply thought of cloud as start small and grow anytime. While I do agree their numbers began small, it is not the realistic expectation to depend on a cloud as there isn't an infinite resources one can obtain which plays really well for the cloud provider to recommend to reserve availability, which is absurd since this somewhat resembles the traditional hosting models where unused resources aren't an issue anymore since you intentionally want it to be there to be used at any point of time.
@@jesteriruka4215 thats always my question. There are many companies offering hosted servers at a fixed monthly rate that wont destroy you with weird fees in the long run. Why are they always ignored in the conversation??
For a lot of use cases, VPSes work well. You'd be surprised at how far you can go with a $60/year VPS (can get a nice VPS with 16GB RAM and 70GB disk space on a 10Gbps connection for that price). Some people will say that you can't scale up as easily... But for the same price as a provider like AWS, you can get at least 3-4x the capacity across multiple regions, and have far more room to grow.
Great analysis on the addictive quality of cloud services and the difficulties in breaking away. I appreciate the options provided to navigate cloud reliance.
I'm lucky I bought a few servers myself and learned the tech fundamentally! I can now already get out of this cycle the moment they start jacking up prices :D
@@Malix_Labs I'm just one guy hahaha! I love understanding the tech I use so I did all the wiring routing researching myself! I set up my server with a static IP and have it first try to route through cloudflare and I gave it a secondary route through google. I made a tunnel with command prompt for Mac and myself he's an awesome editor! I just don't want a megacorporation to hold an entire monopoly on knowledge which is why its critical that I understand the internet from the ground up and learn each of the steps that computer scientists before me used to get us to where we are today! My goal is to show that even an individual still has the ability to compete with the top companies in this world. That while difficult we can relearn and correct our mistakes! I actually independently make AI and to my knowledge was the first person to combine Natural language processing models with prompt based image models! I livestreamed me solving it in 3 hours hehe! I'm going to make videos to inspire hope in people especially for the new generations that they can still carve their own path and create amazing things in this world! That said help would be nice I'm still pretty naïve but very optimistic about what will come in the future :D
The issue of getting them addicted young is real, and insurmountable. Colleges don't create developers or systems engineers in the true sense - they are just turning out people that use package managers, third-party libraries and Kubernetes/Docker. I asked a 10-year software dev to replace the 2-3000 lines of open source code that he imported to implement a freaking hash table cache and he about died. Could not do it. I didn't even ask him to create a hashtable or B+ tree...just take the stock java one and use it smartly. No can do. It's been months and he is still panicked. This is typical of the industry. I'd love to see a Fireship video on all these package managers do to code quality. I once compiled the AWS S3 java library and it d/l'd hundreds of packages for an hour and took another hour to compile. No way that's been audited by anyone. It's an injection attack wet dream.
I do not see anyone escaping the cloud once they are hooked up. Witnessing this rn at my workplace. So many different services and functions with paying tiers. It is incredibly easy to setup for developers. Like kickstarting a big project with no knowledge about the internals. Just watch a youtube tutorial and 1 click in the cloud platform manages 1 -5 hours of configuration for you. The downside: you code only fits this one cloud provider. Especially if we are talking about infrastructure as code. This has 2 side effects: 1) Things randomly stop working because you are hitting tier limits of things you did not even knew they had cost associated limits, 2) the cloud provider now is your personal mafia and can crank up the prices as they like. With your code custom tailored to the cloud provider's platform you are not going anywhere.
I understand all the points and I agree with most of them, people need to be very mindful about how they use these things. But opening your own data center and implementing all the APIs and automation as well as making sure it is reliable, highly available, cross-region, ... . I promise I hate all those companies more than anyone, but if you know what you are doing and you fight BS in your company, you will be able to host things relatively cheaply on those services. Most high-bills issues I've seen are due to the company not caring, rather than AWS having high prices (there are cases for sure that they have outragous pricing/charges). I'm all for open market and competition, but we can't just suggest that the cloud doesn't help, or it is as trivial to open your own data center.
I’m working on my own project now and I was so hesitant about using aws but I’m glad I did it… yeah it’s pricey but learning how to optimize cost and switch gears to use a cdk rather than host it on a pre designed cloud is invaluable.
Cloud only makes sense for small startups, where a server monthly rental has lower up front costs than purchasing your own hardware. But once the profits start rolling in purchasing your own HW is much more cost effective. You don't even have to host it on-prem. You can host at a private data center. Called private cloud.
At the very large end, you are in a better position to negotiate pricing where you can host your stuff on a cloud provider and where you can also negotiate for stuff like dedicated 24/7 support
I switched from the cloud ages ago for my personal server. I bought a $250 Dell r610 used from online. It's not the fastest thing in the world, but perfectly suitable for my needs. Even backing my platform up to aws was getting expensive so I bought a NAS to do the work. Couldn't be happier.
It depends on business model. High margin projects can easily sustain cloud costs while focusing on R&D instead of trying to pass ISO and met clients certification demands with self hosted servers.
This pretty much summed up all my concerns with Big Cloud far better than I could have. I've got a few friends and colleagues already drunk on the Koolaid I'm keen to show this to 😉
Learning all cloud utilities and having to pay for these sounds way to stressful. I was able to take an old pc of mine, put linux on It, get an IP from the internet provider and host everything on It. It's harder, but way more rewarding and cheaper in the long run. The cost of electricity+your time is a better alternative than paying amazon thousands of dollars.
@@vectoralphaSec Fortunately... Or unfortunately... Others were containing some family stuff and some MTV recordings. The movies are called: "Ai No Borei" and "Ai No Korida" And the most ridiculous thing is that it was translated and dubber by Volodarskiy - one of the most famous USSR dubber who worked with Terminator, Rembo, Bladerunner, even Back to the Future
When you set up a new private-public keypair in the cloud, you cannot export the private key "for security reasons". But it also means that you will never be able to import that private key to a different cloud and thus will never be able to sign or decrypt data in the new cloud unless you change your public keys out there.
If you were to write a bot that could manually scrape the data from a server and manually copy it to another server for you, you could escape the egress trap, because you are just accessing the data, not moving it.
(Don't tell anyone about the 4 option: all the code and ̶u̶s̶e̶r̶ ̶p̶e̶r̶s̶o̶n̶a̶l̶ data is in the developer's computer, no extra cost for the company.)
Cloud is cheaper, if your concern about eegres simply manage you own database on your server, and compute on cloud. I recently learned CDK and its so simple to write whole stack in lambda and dynamodb. Using docker is good too, but then you will need to do a lot of devops stuff, while cloud manages itself.
It's all (or at least should be) about expenses and business value. Public cloud providers have out-of-the-box solutions like even mid-cap companies won't ever achieve on their own. For instance, geo-redundant storage services (even on PaaS level like a DB), global private backbone network, highly multi-level and advanced security, and of course the economies of scale. Managers at most companies will also prefer OpEx over CapEx for many (good) reasons. On the other hand, cloud wasn't invented for well predictable static workloads. That's really the case of OnPrem. The primary benefit of the cloud is the ability for scaling -in and -out dynamically, within a short timespan, quasi without limits, and paying only for the resources being used, only while they were being used. Cloud is just yet another tool in IT, like so many others, and can be used well or wrong.
Cloud is for initial days. If business successfull, enough cash flow, start building your own cloud, leaving the geographic scaling part to AWS. The Kafka and web servers can be aws. One can have in house DB and other micro services.
All jokes aside, it actually makes sense to say cloud serivce is not meant to be cheaper but more convinient. It is almost impossible to justify the monstrous bills for cloud service, but it is very convenient to use.
Lock-in is the biggest concern - features that are ONLY available from a particular cloud. I'm never using those unless there is an alternative readily available.
Imagine seeing this and knowing since forever that free demos are free in order to get you to pay. Sometimes i wonder whether humanity has any self-control at all, or whether most people just don't think
As a society we need to aggressively look to keep regulations and observation of the AI/tech space as it continues to make revolutionary changes in our world.
I tried few cloud option and the best one for price is Contabo. BTW all depends how much effort you wanna spend on maintaining your infrastructure. If you need some storage (NAS) the best option is an Hybrid Solution but you need a good connectivity.