- Published on
On-prem vs. Bare Metal Cloud Deployments
- Authors
- Name
- Alex Lee
- @alexjoelee
The year is 2024, and the general population has access to tons of incredibly cool and unique solutions for hosting a websites and web applications. Developers and administrators might choose to keep their backend services separate from their front end services, or host them on completely different providers (e.g. Vercel in the front and GCP in the back).
Looking at cost, though, it's plain to see that you're paying a premium for not having to maintain the systems intricacies. The cloud provider handles the data center and networking contracts (or facilities themselves,) the server hardware, and the server hypervisor/OS. This is worth it for many who focus on software development rather than system administration or operations. If this cost is worth it to you, then stop reading now.
Everyone seems to be talking about going back on-prem... again. These marketing cycles come and go but it seems that a significant portion of SMB's have woken up to what's real: the cloud never saved them all that much money! So, in light of the recent push for going back on-prem: We want to highlight a middle ground that brings far more value than traditional hyperscale clouds: Bare metal! 
Bare Metal
There are countless providers (not always do they call themselves 'cloud' providers,) who offer recent-generation bare metal servers for an attractive monthly fee, bringing upwards of 10x the amount of compute resource per dollar spent. The trouble is managing and maintaining that server's operating system. Like anything, there's a steep learning curve here and your traditional developer should not be thrown into the task blindly. Security is a major concern here: You are responsible for securing that system.
There are other factors to consider when you're thinking of rearchitecting your infrastructure. Who's responsible for the failure of what, and how quickly do they need to respond to you? Having monitoring set up is paramount here, too. When AWS goes down, lots of people know, because there are lots of services hosted in a on AWS. When your bare metal server goes down, there's nobody on it but you to notice.
There's also nobody else but you to fix it, or to engage the remote hands at the datacenter to go check out the issue. There's nobody else to back up your data either, so make sure you do! It's almost certainly faster to restore from a backup or rebuild from a CM than to troubleshoot a broken box. Assess what's most important for your organization. 
On-prem
For those who insist on going on-prem entirely, be prepared for some new challenges compared to the "old days" when you'd stick an old server in a closet and forget it about it for ten years. In most cases these days, downtime is not acceptable, which means you're going to want to think about the space for your equipment. Consider physical access and security, cooling, power, and network. Your goal should be to build a 'mini data center,' rather than a 'server closet.'
Companies today are still deploying on-prem data centers and racks. Some businesses never stopped doing it this way. With the relatively low price of last-generation used hardware, SMB's can benefit from running their own equipment on-premises, provided they consider the up-front security and environmental investment. If your primary needs are private cloud/internal tools, on-prem may even garner you a significant performance increase.
No matter where you host, you can always accelerate your page using a CDN or a Web Acceleration Platform like Skip2. Your web traffic is proxied to our edge network, cached and compressed, and sent to your user. Since our edge network is global and sprawling, users almost everywhere will notice and enjoy a performance increase when visiting your site or app. The features certainly don't stop there, learn more about Skip2 Networks.