Enterprises Turning Toward Bare-Metal Clouds for Critical Applications

Seven Criteria for Choosing an IaaS Provider

Even though we talk about “the cloud” as if it is one giant thing, it is common knowledge that we can, in fact, choose from many clouds. Already, the various SaaS, PaaS and IaaS architectures are starting to align themselves with various vertical markets and select application sets to make it easier for the enterprise to find the right cloud to fit its needs.

But on a more fundamental level, two distinct means to implement a cloud-based infrastructure exist, which in large part parallel the two basic means to provision the data center: physical or virtual. This may seem like a misnomer at first, considering that the cloud probably wouldn’t even exist without virtualization, so why would anyone want to tie themselves down with a bare-metal cloud?

But the fact is there are plenty of reasons, according to tech analyst Ashar Baig. Probably the top one is reliability, which is enhanced by reserving dedicated physical infrastructure on which to build remote cloud architectures. Under the standard virtual scenario, systems are shared across numerous clients, which can lead to resource contention and bottlenecks when the data load from one or more clients starts to spike. Dedicated virtual servers can alleviate this burden somewhat, but a bare-metal cloud is more customizable than a virtual one and can be more easily enveloped behind existing security schemes. The cost of a bare-metal solution is higher, but it is worth it for higher-priority or even mission-critical functions.

It is also becoming easier to integrate bare-metal solutions into broader cloud architectures. Both Intel and AMD have embraced the OpenStack Compute protocol, which allows OpenStack to be provisioned across physical server nodes. AMD recently added OS Compute support to the SeaMicro SM15000 line, which scales up to 512 cores and 5 PB of storage connected through the Freedom fabric that supports up to 160 Gbps throughput.

Many cloud providers, in fact, are starting to play up their ability to provide bare-metal infrastructure as demand for higher levels of service increases. Internap, for example, now provides both virtual and physical OpenStack services under the AgileCloud platform, allowing enterprises to tailor their cloud in a wide variety of cost, security and flexibility configurations. The platform’s bare-metal AgileServers function as a standard instance to enable seamless integration across hybrid cloud architectures. As well, the system supports Layer 2 VLANs to enable hybrid environments across colocation and managed hosting services.

Cloud management platforms are also taking a renewed look at bare-metal infrastructure. OnApp’s new version 3.1, for example, provides automation for traditional servers and what the company calls “smart servers,” which are dedicated devices built around thin KVM layers with hardware passthrough capabilities that can be used for failover, rapid scaling and provisioning functions. In this way, organizations can bring the management of the entire cloud stack, including servers, VPS, DNS, storage and templates under a unified drag-and-drop interface.

Bare-metal clouds, then, are great for taking much of the guesswork out of the cloud. Costs are generally higher than standard virtual instances, but they are more predictable. That and the performance benefits from utilizing dedicated resources make them a likely “next-step” solution for organizations looking to move their cloud architectures beyond bulk storage, recovery and back-office applications.

For enterprises worried about Big Data and the rise of mobile traffic, a bare-metal cloud is probably the closest thing to a brand-new data center without the messy integration and substantial upfront capital costs.