IBM Brings Kubernetes Service To Bare Metal
IBM introduced an option Wednesday to provision its managed Kubernetes service onto bare-metal cloud servers—a first-of-its-kind capability Big Blue claims will drive the development of artificial intelligence applications.
Bare metal, as opposed to virtualized multi-tenant environments, is for many enterprise use cases a superior environment for running data-intensive apps, including those that implement machine learning at scale, said Jason McGee, an IBM vice president and fellow, in a blog post.
IBM's fully managed Kubernetes offering, IBM Cloud Container Service, is the first from any major cloud provider to offer the bare-metal option, McGee said.
"This will widen the potential of Kubernetes, along with the significant agility and flexibility it brings to data, to apps and workloads that require extremely high computing performance, such as large machine learning workloads, as well as sensitive datasets that often require isolated servers," McGee said.
Kubernetes has become close to a de facto industry standard for orchestrating containerized workloads, and just about every hyperscale cloud platform now offers a managed solution for implementing the technology first developed internally at Google.
Companies migrating to the cloud are looking to take advantage of Kubernetes in new ways "tailored to their specific data needs and customized to what works best for certain workloads," McGee said.
New and data-intensive workloads, such as machine learning, require high levels of computing power that dedicated physical servers excel at delivering, he said.
In the past, running Kubernetes required a significant amount of configuration and management from developer teams, limiting its use for production apps. Managed services ease that process with benefits that include automatic updating, intelligent scaling and built-in security.
By extending its managed service to dedicated servers, IBM can deliver Kubernetes in a form that fits any organization's cloud strategy, he said, such as building a cloud-native machine learning app, processing large workloads or migrating apps that ingest large amounts of data.
"This gives developers greater control over where their workloads reside and enables them to isolate workloads to specific servers," McGee said.
IBM is also working with the open source community to make it easier for apps built with Kubernetes to access GPUs, he said. Graphics processors, which excel at crunching numbers, are a "make-or-break component of many cloud-enabled technologies" that are compute-intensive and need advanced performance to be competitive, like machine learning.
"This marks the next evolution in the IBM investment in containers and to our commitment to grow them as a secure, stable and widely adopted component of companies' cloud strategies," McGee said.