Equinix, a provider of hosted IT services, added an Equinix Metal offering that allows IT teams to automate the deployment of bare-metal servers.
Interest in deploying bare-metal servers is increasing due to the rise of both artificial intelligence (AI) applications that are compute-intensive and Kubernetes clusters that, in many cases, don’t require the overhead created by a virtual machine. However, it is more difficult to automate the provisioning of bare-metal servers compared to provisioning virtual machines.
To address that challenge, Equinix earlier this year acquired Packet, provider of an IT automation framework for bare-metal servers that now drives the Equinix Metal service. As part of that service, the framework has been integrated with Equinix Fabric, formerly known as Equinix Cloud Exchange Fabric, to gain access to a software-defined interconnection service spanning 237 data centers deployed near major internet exchanges across 46 geographies.
Jacob Smith, vice president of bare-metal strategy and marketing for Equinix, said those servers are available either on-demand or as reserved instances that can be spun up in less than 15 minutes, similar to the way IT organizations provision cloud resources. The servers can be provisioned via application programming interfaces (APIs) common libraries or employ any one of several tools that they already use to manage infrastructure-as-code, he noted.
As applications become more distributed, latency issues that arise because of network physics are becoming more challenging. Many IT teams are now trying to deploy applications closer to where data is created and consumed. Equinix Metal enables IT teams to achieve that goal in an IT environment that also gives them their own private internet managed by Equinix, said Smith.
At the same, Equinix Metal reduces the time it takes IT teams to deploy bare-metal servers in a rival co-location facility to minutes from days, via an automated service, he added. Not many IT organizations are interested in dispatching IT personnel to physically provision servers in the wake of the COVID-19 pandemic. The health risks involved in traveling to a co-location facility to configure hardware are deemed too high.
In the meantime, as more application workloads find their way to the edge, the physical locations of servers start to matter a lot more. It may now only be a matter of time before IT teams rediscover the merits of hosting services that can provide as much flexibility as a cloud service provider with far less adverse impact on application performance.
Of course, that also means many existing DevOps processes will need to be extended to include not just a local data center or cloud service, but also a network of data centers that are being managed by someone else on their behalf.