It is week seven of Google's OnAir videogasm and that means – if anyone is still paying attention – the latest topic is application modernization, which in Googleland is code for (you guessed it) running your stuff in containers, preferably via the bundle of services it calls Anthos.
The cloud giant took the OnAir opportunity to introduce several new Anthos features. These include:
- Anthos attached clusters, which lets you attach clusters running on Amazon's EKS (Elastic Kubernetes Service) or Microsoft's AKS (Azure Kubernetes Service) to your Anthos estate, including the ability to configure them with Anthos Config Management and to enable a subset of Anthos features on them, specifically Anthos Service Mesh, Config Connector and Policy Controller. EKS or AKS clusters must be running Kubernetes 1.16.
- Enhancements to the Cloud Code plugins for VS Code and IntelliJ IDEs. The focus is on developing applications for Google Cloud Run, which is its serverless container platform. There is a new Cloud Run emulator that runs locally for test and debug. The supported languages are Java, Node.js, Python, and Go. C# and .NET developers should look elsewhere. Devs can also now create Kubernetes clusters from within the IDE. To avoid bill shock, we recommend careful use of this feature.
- Anthos Identity Service is a new feature with support for OpenID Connect that extends an existing directory to work with Anthos workloads. This is available now for on-prem Anthos, and in beta for Anthos on AWS.
- New Anthos migration features let you convert VMs running on-premises to Anthos, run Windows containers (in beta), and monitor migrations in the GCP Web Admin console.
- Anthos for Bare Metal, first introduced in May, is an option to run Anthos directly on Linux without a hypervisor layer. Supported operating systems are CentOS, Red Hat Enterprise Linux, and Ubuntu. This service, aimed mainly at edge deployments, is now in beta.
- Hybrid AI for Anthos lets you run Google's AI services on-premises, with the first example being a speech-to-text service.
But what is Anthos? The word means "wild flower" in Greek, and if that seems an opaque way to describe a subscription bundle, it is doing its job. From a billing perspective, it is something you currently pay $9.00 per vCPU for pay-as-you-go, or $6.00 per vCPU with a committed subscription, unless it is on-premises where the cost leaps to $75.00 or $50.00 respectively. For your money you get the features listed here, which differ depending whether you are on-premises, on AWS, on Google Cloud, or using the new attached clusters.
Still confused? You could dive into this technical overview and try to make sense of the current diagram, showing Anthos as a complete platform with GKE providing the low-level infrastructure (except in the cases where it does not), and then on top of that the cluster management, service mesh, config and policy management, application deployment mechanisms, and at the top developers slogging away writing code.
Google has also come up with a paper called "Anthos under the hood", which is one of the company's better efforts at picking apart its wild flower. "Anthos runs on an infrastructure layer that's been abstracted, and applications that run on it have access to high-value services," we are told. Note that Cloud Run is a key part of the stack, and from a DevOps perspective Cloud Code, Cloud Build, and then Cloud Run is perhaps the ideal Anthos workflow.
The key question, perhaps, is what added value you get from Anthos over and above simply using the GCP services you require, including Cloud Code, Cloud Build, and Cloud Run. We think Google could do a better job of making this clear. ®