audio
As it stands, if you're a business that needs hosting, there's always a tradeoff.
If you want unlimited scalability, you can get that with hyperscale cloud, but it comes with added expense, possible performance degradation, and a lack of support.
If you want reliable performance and control over your infrastructure, you can get that with bare metal, but you'll have to compromise on instant scalability.
The truth is, most businesses need all the above at some point. And it is why more organizations are adopting hybrid infrastructure strategies that combine the power of bare metal and hyperscale cloud infrastructure.
We're strong advocates for multi-vendor solutions, tapping into the benefits of each infrastructure type to find the sweet spot. But we're also of the view that infrastructure as a service (IaaS) can go further to solve this dilemma faced by thousands of organizations.
It's exactly why we're introduced our new offering – hybrid bare metal cloud.
In this blog post we'll explore what hybrid bare metal cloud is and how it works.
toc
Hybrid bare metal cloud is a new type of hybrid infrastructure deployment based 100% on bare metal server hosting. Launched by servers.com in _____, it involves combining different types of bare metal compute on one private network.
There are two key differences between a hybrid bare metal cloud deployment and a traditional hybrid infrastructure deployment. The first is the variety of compute types being combined and the second is the mix of infrastructure providers used to access this compute.
Traditional hybrid deployments involve combining multiple compute types across different hosting providers. For example, a business might choose bare metal hosting or on-premises deployments for their predictable workloads and choose to spin up virtual machines (VMs) from a hyperscale cloud provider during periods of increased demand. In some cases, they may also run GPU-based AI workloads with another provider.
By contrast, our hybrid bare metal cloud model enables businesses to host all their workloads, regardless of type, on servers.com dedicated hardware. So, one vendor, one network.
The hybrid element comes from the ability to combine three different bare metal products that are designed to accommodate three specific workload types: scalable workloads, persistent workloads and AI workloads.
This reduces dependency on hyperscale cloud providers which, until now, have been the only viable option for accommodating unpredictable infrastructure scaling requirements. At the same time, it also enables businesses to bypass the operational complexities that come with managing traditional hybrid models.
|
Parameter |
Traditional hybrid infrastructure |
Hybrid Bare Metal Cloud infrastructure |
|
Compute type(s) |
Any combination of hyperscale cloud, colocation, bare metal hosting and/or on-premises infrastructure. |
Workload-specific bare metal compute solutions, combined as part of a single solution. |
|
Network usage |
Infrastructure deployed on multiple networks. |
Infrastructure deployed on one network. |
|
Vendors |
Requires managing infrastructure across multiple vendors. |
All compute needs met through one vendor (with option to tap into third-party vendors if preferred). |
Combining Enterprise Bare Metal (EBM), Scalable Bare Metal (SBM) and AI Compute (AIC) in one place and on one network, businesses can run high-compute workloads without the need for hyperscale cloud or other third-party vendors.
Unreliable performance, rising costs and lack of customization in hyperscale cloud environments are three of the biggest factors pushing businesses towards bare metal hosting solutions. EBM is designed to eliminate these risk factors, to deliver a combination of reliable performance, cost optimization and customization.
EBM is short-to-long-term bare metal hosting built on a foundation of high-performance, dedicated servers. With no virtualization process involved at any point, EBM users maintain sole tenancy over their bare metal server for the duration of use.
There's also the option to customize RAM and storage or create bespoke setups on a long-term agreement. This helps businesses optimize their hardware at a granular level and many find they can also reduce spend as a result.
EBM is built to accommodate persistent workloads and can be scaled alongside predictable increases in demand. It's ideal for businesses looking to optimize their steady baseline compute requirements short-to-long term because it enables you to tailor your technology and achieve greater performance and cost efficiency.
Imagine you're a live sports streaming platform. You have consistent viewership through the year, but there's one annual sporting event that always causes viewership to spike exponentially. In the past this has overrun your server capacity, causing downtime and viewer dissatisfaction. You resort to bursting into hyperscale cloud but the expense to do so is significant.
That's where EBM can help.
EBM enables you to create a highly efficient and performant underlying server infrastructure and plan for predictable scaling events when you'll need additional compute. Now your sports streaming platform never has to worry about reputation-damaging downtime or incur a hyperscale cloud premium.
Lack of scalability is one of the biggest barriers to entry for businesses considering bare metal. Leaning on hyperscale cloud, entirely or in part, has long been the only way to access scalability on-demand. SBM is designed to fill that gap by providing truly scalable bare metal hosting.
At the hardware level, SBM is just as powerful as EBM because the underlying technology consists of the same high-performance, single tenanted dedicated servers. There's no virtualization process involved, just pure bare metal hardware.
What makes SBM different from typical bare metal hosting solutions is that it's highly scalable – and this scalability comes down to how the service is delivered.
Our SBM servers are available in preconfigured flavors with inventory kept in stock on reserve. Because of this, SBM servers are ready to deploy in a matter of minutes, billed by the hour and spun down again just as easily. In simple terms its dedicated, single tenanted bare metal machines that you can scale up and down.
SBM is built to accommodate short-term workloads, making it ideal for businesses in need of temporary bursts of scalability. When demand spikes you can easily scale up your persistent workloads using SBM then spin back down once demand stabilizes.
Imagine you're a game studio. You launched a new title three months ago and now player numbers have stabilized. You've got steady player numbers that you can easily accommodate on month-to-month enterprise dedicated servers with EBM.
But you also know there's a myriad of factors that can influence player numbers. Anything from releasing new DLCs or patches to the influence of online streamers could result in a wave of new players. So you still need the option to scale your infrastructure fast.
That's where SBM can help. You can tap into SBM any time and top up your compute on an hourly basis until player numbers stabilize again – just like you would with hyperscale cloud.
EBM and SBM are designed to work together. When combined they create a hybrid bare metal cloud setup where you can manage your persistent and short-term workloads on one network.
Host your persistent workloads on EBM then burst into SBM when demand spikes hit. And if demand stabilizes at a higher level than your previous baseline, convert your SBM server into an EBM server to use long-term without the on-demand premium.
Fueled by AI adoption, public cloud spending is projected to reach $723 billion in 2025. It may be the infrastructure of choice for near-instant GPU capability, but it can get expensive fast. That's the point at which AIC can step in to help optimize those workloads for long term rollouts and reduce costs in the process.
AIC is our service for bare metal GPU servers. In short, it is our bare metal offering optimized for AI workloads. The key point of difference between our AIC service and other GPU hosting options available on the market is that it's delivered on dedicated servers (not virtual machines) and over long-term agreements.
This means there's no multi-tenant overhead or on-demand premium, so it's significantly more cost effective compared to hosting GPU workloads in hyperscale cloud environments.
AIC is designed to support businesses that are looking to roll out AI workloads long term on highly optimized infrastructure. To put that into context, if you're a business still in your AI discovery phase and you want to spin up a GPU for a couple of hours to test out a new language model, AIC isn't going to be right for you.
Conversely, if you're a business that has already been through the discovery phase and you're looking for ways to optimize your AI workloads long term, then AIC is the ideal match. You'll be able to build a custom architecture tailored to your CPU, RAM, GPU, storage and networking requirements and based on your specific training pipelines, model size and data bandwidth needs.
Most businesses that require GPUs also need CPU processors for a portion of their workloads. So, whilst AIC can be used as a standalone product, it's designed to work seamlessly alongside SBM and EBM. It means instead of managing your CPU and GPU requirements across different environments, you can keep everything on one network.
|
Parameter |
EBM |
SBM |
AIC |
|
Scalability |
Mid |
High |
Low |
|
Customization |
High |
Low |
High |
|
Billing period |
Short-to-long term |
1 hour |
Long term commit |
|
Best for |
Optimizing for predictable workloads |
Fast scaling for temporary demand spikes |
Optimizing for long-term AI workloads |
With hybrid bare metal cloud, we're offering businesses a way to maintain exceptional performance, reduce costs and scale at the same time. In other words, a way to eliminate the infrastructure tradeoff that organizations have been working around for years.
By introducing scalable bare metal hosting via SBM and combining it with highly customizable enterprise bare metal (EBM) and AI hosting (AIC), we've created a suite of interoperable bare metal solutions capable of hosting varied workload requirements in one place.
At its core, hybrid bare metal cloud is about simplicity. A way to manage variable workloads without having to juggle multiple infrastructure providers. And, more importantly, a way to scale without suffering hyperscale cloud's ever-rising costs and complexities.
If you're interested in learning more about our hybrid bare metal cloud, get in touch with our experts, we'd love to talk to you.
