Data Center Knowledge is part of the Informa Tech Division of Informa PLC
This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC’s registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.
Christopher Tozzi | Nov 05, 2021
In some senses, “bare-metal server” sounds like an outdated term. The days when most workloads ran on bare metal were the days of monolithic applications and on-premises server rooms. That era ended at least a decade ago for most businesses, which have since moved to the cloud.
And yet, returning workloads to bare-metal servers is becoming an increasingly attractive proposition in some scenarios – especially when those servers are hosted in colocation facilities, rather than on-prem. Indeed, by offering easy-to-deploy and easy-to-scale bare-metal infrastructure, colocation providers are carving out a niche that gives them an advantage over public clouds and on-premises data centers alike.
Here’s how bare-metal colocation works and why it is becoming more and more important in the cloud-native age.
Bare-metal colocation is the deployment of workloads directly on servers hosted in colocation facilities. There is no virtualization involved.
You could always deploy bare-metal workloads in a colocation center, of course. But, as noted above, doing so has felt anachronistic by some measures. The standard approach was to deploy applications in virtual machines, with the virtual machines hosted on bare metal. This strategy made it easier to increase security by isolating workloads. It also simplified the migration of workloads from one server to another.
Of course, the big downside of virtualization is that applications running in VMs can’t directly access the underlying physical hardware – at least not without relying on some special tricks within the hypervisor to expose bare-metal resources to virtual machines, and even then, performance can be flaky.
Bare-metal colocation (and bare-metal infrastructure in general) solves this issue by allowing organizations to deploy workloads directly on physical servers, where there is no hypervisor separating them from the hardware they need to access. That setup is particularly advantageous for workloads that are designed to take advantage of special hardware functionality, like the use of GPUs to crunch large volumes of numbers quickly.
It’s not only in colocation centers that you can deploy bare-metal workloads, of course. Public cloud providers also offer bare-metal server options. Indeed, vendors like AWS and Google Cloud have been increasing their lineup of bare-metal instances in recent years.
Bare metal in the public cloud is extremely easy to deploy because you can manage it using the same tools that you’d use to administer cloud VMs. You also pay only for what you use, typically based on the total hours that your bare-metal cloud servers are running.
So, why would you go to the hassle of running bare-metal servers in colocation centers when you can just spin up bare-metal instances in your favorite public cloud?
There are a few key reasons:
The last advantage is probably the biggest reason why bare-metal colocation is becoming more and more common. The ability to deploy bare-metal servers in a colocation center of your choosing, using the same seamless experience you expect from a public cloud, gives businesses the best of both worlds: the control and security that come with colocation, and the speed and scalability that come with public cloud.
To be sure, not every workload needs to run on bare metal. But for those that do, colocated bare-metal servers are an increasingly attractive option. Although public clouds are also competing for this business, colocation providers currently enjoy the upper hand, given the limitations that cloud vendors place on the bare-metal instances they offer.
More information about text formats