Once Upon a Time I used to run both the public and private cloud teams for an organisation of a few thousand (servicing a few tens of thousands of customers). That experience pretty much sold me on the fact that “private clouds” are basically just a giant waste of time and money.
It wasn’t even really the private cloud vendor’s fault:1 we were using Microsoft’s then on-prem Azure-ish/”fabric” service, and it in-and-of-itself was fine. It was the freakin’ hardware that sucked. Which was both, a) expensive, and b) kept breaking due to, specifically, some weird firmware issue between the blades and the SAN that neither hardware vendor could fix. We never did manage to get reliable uptime, and any potential cost gains were obliterated by the amount of FTE my guys put into troubleshooting and hand-holding hapless vendor techs.
Meanwhile, our public AWS deployment was getting something like 20% month-on-month growth, all due to organic adoption from business areas. We didn’t really push it, but there was an automated self-service provisioning form on the intranet and, well, people liked getting their shit deployed in twenty minutes… as opposed to the three-to-six months it would take the legacy server team to deploy non-cloud VMs manually.2
Of course, we were in an industry where both, a) the use of public cloud was still “edgy”, and b) our internal legacy virtualization team were fighting against the private deployment,3 meaning I was constantly having to wrestle with cost modelling to justify both adoptions. Unfortunately for me, I kept doing the cost modelling accurately, i.e. by including internal FTE as well as hardware and licencing costs, and the numbers… did not make me popular, let’s say. Firstly because both models were more cost-efficient than the existing virtual model, but mostly because public cloud was so much cheaper than either.
In the end, I left that organisation, in no small part because a restructure would’ve seen my boss take both cloud teams away, because you can’t have a (comparatively) young woman handling any Hot IT Projects, I guess. Last I’d heard, the dude who’d taken over had reverted things back to what he knew, i.e. manually deployed tin-heavy rack-and-stack. Because new things are scary, I guess, even in tech.
Tl;dr, private clouds suck. Stop building them.
- In this one instance. I’ve certainly seen plenty of other private “cloud-in-a-box” style solutions that were… [↩]
- Yes. VMs. On existing infrastructure. Three to six months. I am not making this up. [↩]
- Partly because they just… did not get the difference between the two models. Which, yanno. They’re 100% not alone in, even today. [↩]