>You want to waste money? Hire a car, with a driver, when you need it.

> Want to save money. Learn to drive.

Oh please. As if learning to drive is the end of expenses. If you finance a car, you have monthly payments. If you don't, then you have periodic recurring maintenance bills. You always have fuel charges. You always have insurance charges. You periodically have parking charges.

I know how to drive, but do not own a car. From time to time, I hire a car, but it no where gets close to costing me the amount of owning a car would.

This is probably true in the states where it is insanely cheap to rent cars, but not necessarily everywhere. And even there...

... I run a junker. That is to say a car that will go the dump as soon as it requires any significant expenditure, and the combination saving of not having to finance it, and most years minimal or no repairs, and only needing third party insurance makes it significantly cheaper than renting.

depends on how frequently you need a car. I drive maybe 3 times a year. that's ~2k in gas/rent fees. that's less than insurance assuming each rental is for a week. never mind the cost of actually buying the car.

in practice I spend maybe 1k every year for cars. primarily for vacation. which owning a car wouldn't absolve me from spending.

very well put, i also rent cars because i rarely need them.

but i think where this analogy breaks down is that if i run a service, no matter how many users, peaky or not, at least 1 server always needs to be on, not "from time to time".

> People think baremetal is hard. It isn't. It does take knowledge.

This.

I always wonder how much of the "clouds" success (economic, that is) would have materialized, if the marketing term never got traction, and everyone just called it what it really is: "renting someone elses hardware without physical access, and less, if any, control over how the stack works from the metal up".

In the good 'ol days, when people wanted to put a service online, they rented the racks at a colo, and either stuffed their own hardware in or, worst case, used rented hardware.

Did that require some basic familiarity with hardware? Yes it did. Did people need to know how to setup, configure and administrate a LAMP stack? Sure. Was it guarded against sudden loadspikes by god-knows-how-many layers of abstraction? Nope.

But it worked, and surprise, in 99% of cases, it was perfectly fine if a website ran at sub-optimal speed for a few hours, or went down every now and then.

And the dirty little secret is: It still does, and it still is.

No, no. The costs cloud saves are in staffing and opportunity costs. Everyone knows that it is more expensive than a comparable server, but...it is easy, standard, and available. If you want to a) not have real estate capex, b) not worry about the core ops part of your applications, and c) used to outsource infrastructure to a managed service provider anyway, then Cloud is a viable value prop. Plus, the more of the services you use, the more you app stack becomes "standard" parts with glue code. This is maybe an improvement at the large enterprise scale where home-built apps don't have a reputation for being future proof.

And better credibility when you say “Our vulnerability was on AWS and configuration is hard, but at least we had the default VPC config” rather than “We maintained our own stack and being sysadmin is hard, and the port was exposed on the web.”

Modern cloud services such as S3 or let's say MongoDB seem to have a lot more security footguns than old-school bare-metal. An S3 bucket misconfiguration exposes your data to everyone even if there was never a reason for that data to ever be exposed to the outside world. On a bare-metal, chuck it in a directory outside your web root and someone will have to actually breach the server before they can steal the data.

> but...it is easy, standard, and available.

So is a LAMP stack on a dedicated machine.

> Plus, the more of the services you use,

The thing is, most webapps don't use a lot of services. Backend-Logic in whateverlanguage, a database, and a webserver. Maybe hooked up to some CRM system. That's it for 99/100 webservices.

Yes, the services cloud providers offer are amazing, they are complex, and it is natural for developers to be fascinated by complex things (I know it is for me). But it's important to realize when simple is simply enough.

You're correct on staffing savings, but not on Ops savings.

I delved into this pretty thoroughly last month - https://medium.com/@rykrk/everything-is-just-build-vs-buy-d7...

Except in what minute of my day am I supposed to take off the hat I'm currently wearing to put on my IT Server Room hat? I don't have time to wrangle this stuff any more. I have multiple clients, I have side hustles, I have what's left of a social life after pandemic, I have family obligations. There are only so many hours in a day. If my time become more effecient by throwing a bit of money at the problem, then it is worth it to pay "experts" at something to relieve me of the burden.

There is a difference between freelancers/one person companies and big cooperations though.

Honestly, the question you need to ask in regards to cloud is a relatively simple one: Can I hire a sysadmin for cheaper than using the cloud?

The answer to that, once you start using enough resources, is more often than not yes.

Sure, it takes a while to get to that point, but eventually you will reach break even and it would be cheaper to do it yourself/have your employee do it.

Yes, but these threads of "in house is cheaper than cloud" never qualify at what size company, at what revenue being generated, etc before their version of an answer is true.

I have been on both sides. Large media production companies with very large amounts of fast and redundant storage located on-prem. These range from local attached RAIDs to large shared SAN pools. Their clients also tend to be the types that sue the crap out of you if any of their content is seen by people outside their control. Switching to cloud solutions was (still is) a huge uphill battle. However, the cloud storage needs are no where near the same (not editing content from s3), but storing approved masters for distribution totally makes sense for cloud. Now that the content is in the cloud, why not perform actions on that content in the cloud. Faster deployment, better equipment, blah blah. Next thing you know your entire workflow past editorial is in the cloud. You start to analyze your expenses and compare them to on-prem amortized expenses and you see that it could be cheaper on-prem. Also, take into consideration how long it takes to bring up that new data center. You also have to look at bandwidth expenses. Bandwidth to a new site not directly on the backbone tends to be expensive for non-residential connections. The additional power expenses of that new equipment plus the cooling is also a new expense. Power redundancy you ask? $$$ Now, you need that sysadmin and possibly a small team. At that point, you go back to your cloud rep, and renegotiate fees. You have now created an entirely new department at your company on managing the on-prem.

even worse, i bet a lot are in the situation of "i have cloud AND i need to hire a system administrator".

this is something easily left out of discussions. it doesn't matter if the equipment is in the cloud or on-prem. someone still needs to be able to manage it all. whether they phsically install new hardware or push a button on a UI to bring up a new machine, it is still needed to be done and managed.

Re: Cloud. Not all cloud scenarios are the same. If the cost is amortized over a long time (theoretically infinite, well that's the plan) then the immediate convenience can outweigh the cost/opportunity cost. For example if you used Backblaze to backup one personal computer at the cost of $6/mo, if you have a lot of data that becomes a huge source of savings compared to managing the backups yourself. At that price the ROI versus other methods like building a trueNAS may not be within a decade, and I'd argue the storage enthusiasts have probably refreshed all their drives within that time and the ROI would never come even if Backblaze doubled their prices. What you do get is that self hosting becomes a hobby, and that's what I feel it is for most people.

Hiring a personal car is more expensive because you are hiring a personal employee.

That said, I still argue for personal autonomy alone learning to do the thing is better in general, but I don't think it's because it's cheaper in all scenarios. And to your point some or maybe even most cloud services are more expensive relative to their self hosted versions.

There are savings, but they require work to realize.

Let's use your driving example (because car examples are always great!)...

>You want to waste money? Hire a car, with a driver, when you need it.

>Want to save money. Learn to drive.

This is true. You can save more money if you need to drive often if you own a car. But there are two scenarios that it still makes sense to rent.

1) What if you need a car in a different city? You just flew from JFK to SFO. You already have a car in NYC, but need one in SF. You're not going to buy a car in SF that you'll need to sell in a week. Sure, if you're going to be there longer, you might consider it, but then you're still carrying the costs of two cars.

2) Sometimes you need a truck. Maybe you have an IKEA run to make to get a bunch of desks, or stop at the hardware store for a few dozen bags of mulch, or ... But sometimes you just need a truck to get the job done. You could just buy a truck and be done with it. But trucks can be more expensive than a compact car, and they definitely have higher fuel costs. In this case, you'd probably be better off with a fuel efficient (or electric) compact car and rent a truck only when you need it.

This is how you save money with the cloud. But you definitely don't save money when you effectively rent a truck to drive to work everyday (even if you are in construction). There is a cost to renting -- it is more expensive on a per-use basis than it is if you buy. Cloud servers are more expensive than bare metal -- if you're constantly using them. It is only cheaper when you stop paying for the parts you don't need. And that also takes expertise.

Once, at a new job, I inherited a cloud server. It was costing us a ton of money per month and running 24/7 because the person who set it up never turned it off. After 3 months of those costs, they could have bought a new server with no other renting. They paid for a cloud server for three reasons: 1) they had no experience with hardware, 2) it was a pain to setup local hosting, and 3) it was faster to get running without waiting for a vendor to build a server, deliver it to the datacenter, etc... These were real impediments to the first person and the cloud server helped to get them moving. They just didn't have the longer term view of what their decision was going to cost in the long term.

The first thing I did was order a new server and make friends with our datacenter ops people. And now the only thing we really use the cloud for is archival (write-once, read-never) storage. If we ever really need these data, it will be super expensive. But, if that ends up happening, we'd be happy to pay the cloud tax.

> This is how you save money with the cloud. But you definitely don't save money when you effectively rent a truck to drive to work everyday

Isn't that exactly how companies use the cloud? Sure, there are contrived examples where the cloud is cheaper than self hosting. But the common case is that companies "use the cloud" by putting 100% of their infrastructure and hosted products in the cloud. That's what is meant when you say "X uses the cloud".

I think you're taking his analogy further than he intended then are arguing against your version of his analogy.

Cloud was made for people who don't have the time, talent or desire to build and manage it in-house. You pay a premium for that convenience and that premium scales with your business growth via IT resource needs. I think that's what he was getting at in his analogy.

sure - but then also think about the use case - he is using a storage account, which means that inclusive in the cost is

a - replication (within region / AZ at least) b - 0 software to maintain (no need to frantically patch apache / SSL / whatever) c - super quick set up / management / logs / etc

So, yes, bare metal is (on a cpu cycle to cpu cycle / GB RAM/HDD/Bandwidth) level cheaper, but TCO can be waaaaaayyy higher.

Yes, TCO can be higher, depending where you are on the curve of capex, amortization, and staffing costs. Don't forget you still need at least Developers, DevOps, and Security. If you're inefficient at cloud, spinning up ec2s left and right, using a lot of egress, storing a lot of hot/live data, your total cost is much higher, and will easily be more than the salary of that one sysadmin, or team of systems engineers, you would pay to maintain the colo space.

You have to do a lot of things right to get that Cloud Value, as the author of this blog post has shown. You have to do a lot of things right to get value out of on-prem bare metal as well, but those things are generally well-known, standardized, have less moving parts, and people with decades of experience and knowledge of best practices. The opposite of the current cloud landscape.

TCO is not a straight line.

It's the "weaving" part that has non-specific cost. If you have skill at weaving together pieces of the cloud in an optimal way, you can save money. Just like if you have skill in putting together your own infrastructure you can save money. I can see spending money on services, but I don't understand why people invest brain capacity on vendor-specific solutions.

How much extra time & development effort does it take to weave the services the right way to realize those savings, as opposed to doing it the simpler, old-school way on bare-metal?

In my bare-metal-hosted projects I can afford to do a lot of things that would be a major no-no in the cloud because I have so much hardware resources I can save development time doing things inefficiently and still come out ahead in terms of costs.

This is such a bad and US-centric example: for anyone who lives in a place where it is easy to get around without a car, hiring a car only when you need it is a no-brainer financially and owning a car is a totally waste of money.

No, there are economies of scale. For $5/month I can get. dedicated IP address in the cloud. For me to get one myself I’d probably have to buy real estate somewhere just for starters.