I do often wonder if people will eventually migrate back to on prem
this post was submitted on 12 Mar 2025
166 points (99.4% liked)
Shitty sysadmin
171 readers
231 users here now
Welcome to Shitty Admin!
This community is for shit posting about terrible things computer admin do. Cross posts are welcome and so is satire.
Please no politics or harmful content
founded 4 months ago
MODERATORS
We migrated a bunch of clients back when we took over for other IT. Cloud was slower, way more costly, less utilitarian, and gave less control. I have no idea why people switched in the first place.
I actually brought it up on an MSP subreddit back when I was still posting there and was relentlessly shit on.
Because "cloud" was the hot buzzword of 2005
It is good for some things like a web server but bad for anything that involves high data or compute needs.
The theory is that on-prem includes a lot of ancillary costs like a team of staff for maintenance (or cost for outsourcing it), hardware maintenance/upgrades, cybersecurity, dealing with failures, backup, load balancing, multi-region/multizone etc.
I don't think cloud solves all these issues necessarily and I am convinced if you do the calculations cloud ends up being more expensive depending on the scale. I think you really pay the premium for convenience, speed (of getting things going) and user experience (the software)
The biggest reason I think is SLAs and the ability to blame someone else when something goes wrong. I've seen it play out at multiple different companies now.
SLA and support is the biggest one. We had to pay Migrants for Docker Enterprise licenses just in case we needed support or some sort of liability shield.
Yeah, having to have someone „on site“ who knows about cyber security and such vs having a piece of paper laying around that tells you that availability, continuity, security are hidden away in a SEP field. It’s easy to guess which one you want to choose…
Depends whether you want to try preventing damage, or just have someone maybe* half-ass it and take the blame.
*Like they say in security, "how do you know?".
"Repatriation of workload" is a big thing right now. So yeah.
Hybrid cloud and multi cloud has been a buzzword for a bit.
Colocation or in-housing with your own clusters is slowly getting more practical, as long as you're not dependent on a bunch of vendor specific APIs. Much less effort to port things that are already in containers.
Broadcom fucked a lot of people with the VMware debacle. If I was a business owner I would think twice about where I'm putting my eggs.
Sorry, I've been out of the game for a while. Can you expand upon this? I'd like to better understand.
Qualcomm bought VM ware after Dell spit it out.
They proceeded to basically jack the price and otherwise alienate much of the user base. In a world where AWS is a thing, you saw a lot of companies walk, and there was a general lamentation for enterprise on-prem virtualization options.
Hybrid cloud is where it is at
My workplace started using “onprem cloud” and I can’t even begin to describe how I feel on the matter.
If it convinces the beancounters and management to accept it, it's a win.
I think that’s called fog
Marketing works, and executives aren't that smart
Well there is a distinction. If the on prem hardware is running a cloud stack, which hosts all the actual stuff the company wants, then it's on-prem cloud. The purpose of this is usually to make infra management more infra-as-code friendly, and keep most of the benefits of cloud (convenience mostly) with less cost. Of course it requires hiring people who know how all of that works, something you get "included in the cost" when you pay for off-prem cloud.
Unless you're paying someone run hybrid cloud for you (IBM and other so this), but that's usually for sensitive data or if all your users are in house and sending data to the cloud just to get it right back doesn't make financial sense.
We've on-premed a lot of things in the past year, and plan doing more.
- ConnectWise Screen Connect
- Tactical RMM
- Wazuh
- Matomo Analytics
- XWiki
Whith US uncertainty talk has already started, at least here in Norway more and more are talking about looking at exit strategy. Costs are huge because reinvestment but disruption is worse.
Customer was US government and decision was made by DOGE?
Can't possibly be slower than GCP, either. Performance of BigQuery is something to behold.
Cloud makes sense if you've a very 'spiky' load, I suppose. A website that needs one VM most of the year, but a hundred on a couple of days. Maybe your data processing needs 100 TB of storage a couple of times a month, but not the rest of the time. But for fixed, predictable bread-and-butter stuff?
Save them even more money by taking their desktops all off of Windows and downloading Ubuntu, it's free and I've been running it for over a week with no problems at home!
A whole week!?
Sometimes whole days go by where I don’t have to spend hours fixing a software issue in terminal before I can do my 5 min job.
So easy grandma could do it.
And while you're at it, do they really need backups? I mean, it takes up so much space... And UPS... They need to be lifecycled so consistently, it's such a pain in the wallet.
Oof ow my wallet!
Honestly I use ZFS instead of a UPS. It is much cheaper and works just as well.
At some point the number of pools will absorb the outage.
But how much of that half mill goes into your pocket? Surely that money is padding the execs wallets.
Have you priced out big data and compute in GCP?
Sure it's inflated as hell, but Google has a lot of expensive devs to pay
And I am sure the person hosting a NAS at home has a power bill that also needs to get paid.