Leaving the Cloud: Cloud Computing Isn't for Everyone | Hacker News Reader

Leaving the Cloud: Cloud Computing Isn't for Everyone

basecamp.com

36 pointsemperor_11 comments

While such post are always mentioning costs as a benefit, it would be good to mention that cloud costs are operational expenses and can be deducted completely in the year they occurred while owning hardware is a capital expense and can only be partially deducted over a period of time (like 3-5 years?). As far as I know this has some tax saving implications. Yes, on prem might still be cheaper but it would still be good to mention it in such posts.

The conversation is absolutely more nuanced than most of these articles/blurbs usually mention. Beyond the well-defined hard costs, does it benefit the company more to save money on a cloud footprint, but put that same money into management efforts required to handle manual maintanance of bare metal servers?

Can you afford to spend the time to handle fault tolerance and high availability at the infrastructure level, or will that just fall to the wayside once your migration effort passes its original deadline and the business wants to shift to other items? How will you handle auditing without Cloudtrail, or its counterparts? Will you be able to automate OS provisioning, or will you be stuck with adding more staff and increasing lead times?

Let's say you can give a positive answer to all of these. Now, does it benefit the business more to have your team handle all that's required to maintain your own physical infrastructure, or would that time/effort be better spent on things that differentiate the business from its competitors?

We should also remember that Basecamp has built their entire business around this application. They've had time to perfect it and customize it to their exact needs. This is very different from replatforming commercial or open source software that isn't developed in-house. I think there's good reason that most of the time an article of this nature comes out, it's often a SaaS-type offering that's the target of repatriation.

You can just lease hardware from Dell if converting CapEx into OpEx is your main concern. You don't have to go to cloud for that.

That’s interesting indeed. Good to know

I think this is a very interesting discussion. Also, the pure hardware costs are one thing, but in terms of training and hiring people, will it still paint the same picture? Will you be able to detect and patch security threats as fast? What about utilising archive storage to save costs?

But yes there are nice to see multiple perspectives on whether or not to use the cloud. Does there exist a good managed service to provision on-prem infrastructure in an IaC-matter?

Not just the cost, but you are not under the mercy of someone else ToS and you are in a full control of your system, the only justification against on-premises are either you don’t have the proper infrastructure like a reliable internet access or the capital cost upfront.

You are always at the mercy of someone else's ToS, even if you own your hardware.

This works both ways, though. If something goes wrong with an AWS managed service, they are the responsible party. You have someone to sue.

I think what is often missed is that many thought it's a good idea to just move stuff from baremetal to cloud VMs as is and then wondering about the bills.

IMHO cloud starts to really pay off for cloud native workloads on serverless offerings with FinOps as part of the culture. If you don't have cloud native workloads and don't want to develop such systems, then better keep your sysadmins and stay away from the clouds. There's absolutely no reason to ride the hype train, use whatever works best for the particular use case.

It's what I've been saying all along.

Also in almost all cases microservices aren't needed.

Not a new page, from back in July I think

Reply on news.ycombinator.com