This is like the third or fourth article I have read about some company leaving AWS and saving what seems like a ton of money. The thing they have in common is that they don’t talk about security and other more mundane things such as infrastructure maintenance, which is one of many things you pay AWS to not worry about.
Agreed.
With that $230k/year you’d have to hire more engineers to maintain the servers in addition to the normal day to day stuff they would be doing more quickly in AWS.
You’d also have to simply find engineers who are willing to work on that platform. I personally would not. If someone else out there is willing to figure out the details on pxe booting or the ipmi differences across vendors or hacking snmp data from a switch into a modern monitoring stack, good luck to them. Those days are behind me though. I’m never going back.
You don’t get that much “security”, though. You’re still responsible for actually securing all your applications and infrastructure components.
Also, this is a question of scale, just like with any make or buy decision. AWS isn’t magic, it’s just servers. For a small company, it makes sense to buy servers as a service, but at a certain scale, it stops making sense. This is no different from buying screws versus making your own screws.
To the surprise of no one. Cloud services are convenient and easily scalable - Which makes it cost effective for a lot of work loads. But for all work loads? Of course not!
“We saved $xxx by buying our own car instead of paying taxi every time.”.
They barely touched on the time and money spent managing bare metal. I’d imagine that 230k a year is gonna get a big ol dent in it when their assumption of “modern servers make maintenance needs much lower” turns out to be false.
All for not hosting in AWS but I want to see one of these articles put out actual numbers about what it looks like to run. I want the we are two years into this and have learned X, Y, and Z post. And one that isn’t written by DHH
Also, it’s just super easy to waste money in the cloud. That convenience and scalability is a huge double-edged sword. Even if you try to be careful, analyze pricing structures, estimate your spend, optimize your utilization, etc. (which a lot of orgs and engineers don’t), you can get one thing wrong and end up spending way more than you expected anyway.
Given that, even as someone who likes using the cloud a lot more than my prior experiences with on-prem, I can see why a lot of companies would decide to throw their hands up and switch to on-prem anyway. However, I also feel like a lot of orgs could probably find a happier medium if they invested the time to understand, track, and optimize their cloud usage more effectively.
We’ve come full circle.
No mention of software costs, ongoing software and hardware maintenance. There is no cookie cutter answer, it really depends on the specific systems.
They say their monthly op-ex is $5500.
Our monthly operational expenditure (op-ex), which includes power, cooling, energy, and remote hands (although we seldom use this service), is now approximately $5,500.
If you need it fast, or have unpredictable load spikes, then cloud makes sense. If not, then it costs more over time than the on-prem investment,the end.
THIS
those statements should be in relative numbers
Hilarious lmao. 230k is the salary of 2 or 3 ops engineers 😂
I’m always a little surprised that we talk about NFS and Cloud Native, like gross figures without talking about payroll
I’m gonna go against the grain and say that I like that they are open about this even if the results aren’t amazing. It gives us more information about how low the returns of such a migration could be, and also what tradeoffs they made to complete it. They might be trying to spin it as a win, but as long as the information in the post is accurate it’s good enough for me to use as a datapoint.
Yeah this is not the one to celebrate. They lost high availability and redundancy.
