“The target is for us to have our next-generation AIC that we deploy next year to leverage [Kubernetes] as a foundation of our platform,” van Wyk said. “For our future AIC deployments, both the OpenStack and non-OpenStack components that make up our cloud will run on top of Kubernetes. This is not just an experiment.”
Van Wyk noted that historically, AT&T has required components that run on top of its OpenStack-based AIC to be cloud native, but not the underlying technology. He added that the carrier wanted its customers to bring a cloud-native approach to “our OpenStack cloud, but then thought ‘why have we not created our cloud product that way as well?’”
“We expected those [services brought to AIC] to be ephemeral,” van Wyk explained. “We wanted those to be quick and easy to deploy. We wanted to be able to spin those up and destroy them at will.”
Van Wyk broached the subject late last month as part of a blog post. In it, he said the AIC Container Platform (AIC-CP) would rely on Kubernetes for its automated deployment, scaling, and management of containerized applications; use containers as a stand-alone, executable package of software components; and tap the OpenStack-Helm project for deployment, maintenance, and upgrading of OpenStack services.
Van Wyk said that OpenStack-Helm combines Kubernetes, OpenStack, and Helm to create templates – dubbed Helm-Charts – for each OpenStack service. Helm is the package manager, while Charts are the packages. The Charts packages reside within a Kubernetes repository.
“These Helm-Charts are a mechanism for describing how to instantiate the cloud resources using Kubernetes,” van Wyk explained. “They also instruct Kubernetes how to manage the complete lifecycle of the OpenStack cloud infrastructure.”
This allows for a customizable framework for operators and developers, and it allows end-users to deploy and manage operational OpenStack environments.
AT&T currently counts more than 80 AIC “zones” around the world, with plans to have more than 100 zones by year-end. The carrier had originally planned to hit that triple-figure mark at the end of 2016.
Van Wyk touted a number of operational benefits by going with Kubernetes inside of AIC. These included faster deployment, lower costs, greater control, and higher quality.
The speed benefit comes from being able to more quickly develop, test, and spin up infrastructure requirements. Van Wyk said AT&T would also be able to more quickly perform load balancing across its AIC deployments.
Cost benefits come from a smaller footprint due to greater use of virtualization by “an order of magnitude,” and a resulting decrease in the number of servers needed. AT&T expects greater quality from being able to work more closely and rapidly with upstream projects.
“We looked at [containerization] almost two-and-a-half years ago when we were building the current AIC, but at the time felt the technology and support were not mature enough,” van Wyk said, noting the company went with the plan to virtualize the control plane using more established virtual machine technology. “Then we looked at it again at the end of last year with the view of looking forward as well and found that it was ready.”
Van Wyk admitted that while progress has been made, AT&T continues to tap into the upstream community to bolster Kubernetes maturity. Some of these efforts are under a broader “Treasure Map” project, with individual items labeled with nautical-themed names like “Dry Dock” for bare metal provisioning and “Armada” for orchestration of installs. Van Wyk said AT&T has been working with a number of companies, naming Juniper, SK Telecom, and Intel.
Despite the heavier reliance on Kubernetes for AIC, van Wyk said AT&T is not ditching OpenStack.
“We are committed to OpenStack as an open source initiative,” he said, noting “Helm is core to the container efforts and part of that commitment.”
“We could have come up with another way outside of the [OpenStack] community, but did not think that was right,” van Wyk said. “The core of our cloud is OpenStack and will continue to be so.”