Server virtualization took enterprise computing to a whole new level. No longer did powerful and expensive machines have to sit inside data centers consuming valuable power and cooling capacity while being only partially utilized. Using virtualization to slice up their computing power and renting the individual slices out by the hour over the internet—the public cloud business model as we know it today—was another massive jump forward. Not only did it transform the computing industry, it transformed other industries by enabling any modestly resourced small team or individual developer to deploy an application at global scale, spurring new competition and disruptive innovation.
Now, as companies have spent many years using public cloud platforms, it’s become clear that while the platforms work great for the bulk of applications out there, they are not the ideal infrastructure for all workloads. Be it cost, performance, privacy, compliance requirements or another reason, some workloads continue running on dedicated servers, and there have also been a few public examples of prominent companies moving applications from public clouds back on premises.
But the choice of computing infrastructure isn’t binary. It isn’t between cloud and on prem. There are alternatives, including one that combines the benefits of both: dedicated cloud. In this article we will explain what dedicated cloud is and under what circumstances it is a great fit for deploying an application and then provide some examples of the kinds of businesses that stand to benefit the most from this style of infrastructure.
Flexibility and Control
One condition under which architects reach for something other than a multitenant public cloud is when their application has a specific set of hardware and/or software needs. They can get that level of control from a dedicated cloud provider.
Most dedicated cloud providers offer fully managed bare metal servers and a choice of hardware configurations to optimize the infrastructure for the needs of the user’s application. You may also get a detailed view of the hardware components, such as CPU and GPU model numbers and clock speeds, memory capacity and storage and networking configurations.
Traditional public clouds offer virtual machines and no access to their hardware hosts. Designed to handle a wide variety of workloads, this infrastructure can be suboptimal for especially demanding applications.
Another advantage some dedicated cloud providers offer is a selection of storage options. The Equinix dedicated cloud, for example, offers a variety of direct attached storage configs, as well as heavy duty enterprise flash arrays from the likes of Dell, NetApp and Pure, all fully managed and provisioned remotely.
Users with specific software needs, be it a certain OS image, hypervisor or container orchestrator (to run without a hypervisor), have those options when deploying on a dedicated cloud.
Equinix provides a range of natively supported operating systems and the option to install your own OS via custom iPXE scripts executed during provisioning. For example, an organization could create its own distributed hyperconverged infrastructure by installing a hypervisor like Proxmox, Harvester or VMware ESXi, and run Kubernetes clusters and VMs on top of it. It could also just run Kubernetes directly on bare metal!
Traditional virtualized cloud providers have only a limited selection of operating systems their VMs can be provisioned with. The managed Kubernetes services they offer usually run containers on the same VMs they use across their platforms.
The absence of a hypervisor on dedicated cloud servers means more of each machine's resources are available for processing your workload. This is an important advantage for performance- and latency-sensitive applications. Also crucial for this category of applications are disk and network I/O performance, which is superior on bare metal servers. Finally, there’s no risk of performance degradation due to multiple customers’ workloads competing with each other for the host machine’s resources, something that happens in traditional clouds, where different users’ VMs share the same physical computers.
Dedicated cloud is also a natural choice for workloads where privacy and security are paramount. Provisioning single-tenant servers eliminates the risk of lateral attacks that are possible on multitenant clouds, where a compromised VM can affect the rest of the virtual instances that live on the host server. Having the option to connect your dedicated cloud infrastructure to third parties (such as public clouds, SaaS providers or network operators) privately, without traversing the public internet, is another big benefit for workloads where security is at the top of the priority list. Using a dedicated cloud provider with many locations around the globe helps comply with data sovereignty laws.
While there is the obvious cost advantage to simply consolidating workloads on bare metal servers without the performance overhead of hypervisors, Equinix’s dedicated cloud specifically has some unique characteristics that enable considerable cost savings beyond that basic advantage.
The service provides many options for architecting your compute, storage and network in a way that minimizes data egress and data storage costs. You have full control of where and how your data moves across your network. You can, for example, use one of the aforementioned managed flash array options to store your data (a much cheaper storage option than public cloud) and provision a direct, private link to a cloud service that runs analytics on that data, all in the same data center. You can store data in any of the 27 Equinix dedicated cloud metros strategically located close to end users to reduce data transit costs. You can augment long-term reserved bare metal instances, which come at a substantial discount, with on-demand capacity when and where you need the flexibility to burst compute to meet fluctuating demand.
Dedicated Cloud Is for You When…
Following are examples of businesses whose workloads are naturally positioned to benefit from running on dedicated cloud infrastructure.
The inherent performance advantages of bare metal servers are attractive to game developers seeking to provide high-performance gaming experience. The Equinix dedicated cloud provides additional performance advantages by facilitating deployment at the edge to reduce the distance between players and servers, both minimizing network latency and reducing data transfer costs for the developers.
For financial services companies in the electronic trading space, where high performance and low latency are a matter of competitive advantage, the choice between dedicated cloud and virtualized public cloud is critical. In addition to the hardware performance benefits already described, Equinix operates data centers that are home to the major trading engines and market data providers in New York, Chicago, London, Tokyo, Sydney and Singapore. This enables financial services firms to strategically place their servers in close proximity to the markets themselves, ensuring the lowest possible latency for the trades they facilitate. The dedicated cloud option also stands out in this highly regulated sector because of the greater security and privacy safeguards than public cloud alternatives.
For adtech companies low latency is at least as important as it is for electronic trading firms. It determines efficiency and effectiveness of digital advertising ecosystems. The delay between a user action and the corresponding system response impacts the user experience and, consequently, the success of ad campaigns. Meanwhile, fast data transfer is crucial for real-time bidding (RTB) platforms, where milliseconds can decide the fate of an ad auction. Low latency ensures seamless ad delivery, enhancing user engagement and satisfaction. For successful advertisers, publishers and ad exchanges lowering network latency is an ongoing quest. These organizations are always on the lookout for ways to further streamline data transmission and otherwise optimize their ad tech infrastructure.
High Performance Kubernetes Clusters
Running Kubernetes on bare metal can be faster and more efficient than inside VMs, and the absence of a hypervisor isn’t the only reason—or the biggest reason. Cloud VMs and cloud providers’ APIs make provisioning a K8s cluster quick and convenient, and dedicated cloud enables that speed and convenience to bare metal, whether the user is an organization that needs K8s for its own purposes or a business that provides managed Kubernetes to customers.
This use case is where it’s not only important to have all the compute hardware performance you can get but also where that hardware is placed relative to content consumers (to ensure low latency), how it is connected to the “eyeball” ISPs that carry the content to those consumers and how it is connected to the source of the content (which could be a public cloud or the streamer’s self-operated data center). Control of the entire network architecture by the content provider is crucial here and something traditional public clouds offer only to a limited extent.
Using a dedicated cloud like Equinix’s enables the content provider to cache content copies close to a maximum number of consumers in densely populated metros around the world and connect to those last-mile ISPs and internet exchanges over reliable, direct network links, in the same facilities.
Artificial Intelligence and Machine Learning
The potential benefits of dedicated cloud for AI/ML applications fall into three categories: privacy, training performance and inference performance and latency. The performance advantage of bare metal is the same in this case as in the previous examples. Privacy plays a role for organizations that need to train their AI models on private and/or proprietary data, which they simply cannot transfer to a public cloud. Finally, placing inference infrastructure in edge locations lowers latency for end users.
In addition to the aforementioned security benefit of single-tenant infrastructure (no risk of lateral attacks from a compromised “neighbor” VM), the Equinix dedicated cloud enables robust protection by providing access to private interconnection. Architects can control where and when traffic from their private networks egresses to the public internet. They can set up their global networks to conduct all the processing and data transfers over dedicated private connections between their infrastructure and their cloud and network providers and egress to the internet only in specific locations where it is absolutely necessary. These capabilities are valuable for both large global enterprise networks and SaaS companies that provide security services.
The above are just a few examples of the types of businesses for whom dedicated cloud may prove more effective than traditional cloud services hosted on shared infrastructure.
To sum it all up, dedicated cloud provides the benefits of immediacy, scalability, flexibility and geographic reach of traditional public clouds to users whose workloads have specific performance, configuration, networking and security requirements that cannot be met by virtualized multitenant cloud platforms.
Equinix dedicated cloud has the additional advantage of being able to offer private network access to all the major public cloud platforms and network operators in strategic locations around the world, opening the door for architecting sophisticated hybrid and/or multicloud environments at global scale.
Ready to kick the tires?
Sign up and get going today, or request a demo to get a tour from an expert.