(With Dave Bartoletti and John Rymer)

The biggest announcement out of the Google Cloud Next conference in early April was the general availability of its Cloud Services Platform but with a new brand name: Anthos. Anthos promises to help software delivery teams build and manage modern applications across multiple cloud platforms, including on-premises and public cloud. Anthos combines Google Kubernetes Engine (GKE) and GKE On-Prem, Google’s managed Kubernetes public and private versions, and includes Istio service mesh and Config Management for network visibility and to create and enforce deployment policies. Coming next? Anthos on Amazon Web Services (AWS) and Azure. For good measure, Google released betas of Anthos Migrate (to lift and shift virtual machines into containers) and Cloud Run (to support a Knative-powered developer experience).

If this sounds like a more-curated, Kubernetes-based development platform-as-a-service (PaaS) — something closer to Pivotal PKS, plus BOSH, Red Hat OpenShift, or Docker Datacenter, that’s because it is. Google’s fully in the cloud-native dev platform market now, where not only Pivotal/VMware and Red Hat/IBM battle it out but where AWS and Azure offer strong public cloud (and emerging on-premises) platforms, too. As usual, Google goes down the fully open source software path, promising a consistent experience anywhere.

Anthos Is An Opinionated Cloud Platform

Enterprise development shops face a choice when it comes to building modern cloud-native services on Kubernetes-based platforms. They can:

  • Build a platform from scratch. With this approach, an enterprise stitches together a container runtime (e.g., containerd), orchestrator (Kubernetes), service proxy (Envoy), service mesh (Istio or Linkerd), logging (Fluentd), tracing (Jaeger), security, storage, functions, etc., etc. While a “from scratch” approach offers ultimate flexibility, it also means choosing from multiple options and configuring how components work together. It’s not a light lift. And it requires some hard-to-find talent.
  • Build a platform on top of a managed public cloud Kubernetes service. Starting with Azure Kubernetes Service (AKS) or Amazon EKS/Fargate, enterprises can plug in their own CI/CD pipelines and toolsets. Teams will need to build a lot of integrations and automation themselves, but they don’t worry about infrastructure management. Google’s betting that those who’ve already chosen GKE will want a managed version they can take to any cloud. AWS isn’t there yet (EKS may or may not be available on Outposts when they ship), and Kubernetes on Azure Stack is in preview.
  • Adopt a more opinionated container PaaS. The alternative is to adopt an opinionated platform that preconfigures components and abstracts developers and operators of the lower-level details of choice and configuration. In many of our inquiries with enterprise clients, we find them debating the merits of two opinionated platforms: RedHat OpenShift and Pivotal Cloud Foundry (PCF)/Kubernetes service (PKS). Both incorporate Kubernetes and a host of other services that modern software developers need to build and run microservices at scale.

Anthos is the result of Google’s recognition that most enterprise customers are looking for an opinionated platform that makes choices about lower-level implementation details rather than assembling their own platforms from individual parts. And to compete with other opinionated alternatives, Google needs to build an equivalent branded platform that can work on-prem or in the cloud — and not just Google’s public cloud. The big bet? Customers will love Anthos so much that they’ll run it even on their Azure and AWS clusters — basically just using its rival public cloud leaders for infrastructure. When competing with OpenShift or PCF, opinionated, hybrid cloud support is just table stakes to enter the game.

Google Has More Work To Do With Anthos

And that’s part of the problem with Anthos as it exists today. The Anthos team at Google still has work to do to make it more than just a “me too” hybrid cloud platform, including:

  • Getting Anthos up and running on AWS and Azure. Google took pains to differentiate Anthos from the “hardware” approach of AWS Outposts (AWS hardware on-prem, fully managed by AWS, plus AWS services) or Azure Stack (certified hardware platforms required), but Amazon and Microsoft chose those approaches to ensure on-premises service-level agreements and support for a wide range of their cloud services, not only their Kubernetes platforms. Google aims to be the multicloud Kubernetes control plane on any infrastructure, but while ensuring the same experience, security, and performance on-premises and in Google Cloud is one thing, doing so on any infrastructure is another. Anthos cannot yet be a “run anywhere you want to” solution until then.
  • Tighter integration of Cloud Run on Anthos. Cloud Run is Google’s managed compute platform that runs stateless containers that are invocable via HTTP requests. Built on Knative, it’s Kubernetes- and Anthos-ready, but the product integration and go-to market isn’t there yet. Still, it’s a step up from the “bring your own function-as-a-service” integration exercise that enterprise developers who want to run serverless workloads on Kubernetes are faced with today.
  • Consolidation of Cloud Functions and Cloud Run as part of the Anthos platform. Serverless developers now have three Google services as potential deployment targets: Google App Engine, Google Cloud Functions, and the new Cloud Run/Cloud Run on GKE/Knative. Without prescriptive guidance on when to choose and use each service, enterprise developers may prove wary of all of them. It would make sense for Google to consolidate around Cloud Run as a service within Anthos, much like Pivotal Function Service is a component of Pivotal Cloud Foundry.
  • Moving beyond “serverless containers.” Taking an existing workload and dumping it into a container doesn’t make it serverless. It might still be a “big ball of mud” that doesn’t scale and indiscriminately burns resources. Modernizing monoliths for serverless involves teasing them apart, making individual services stateless and capable of scaling to zero. No amount of “lift and shift” migration magic will get enterprise developers around the technical debt embedded in these monoliths.

Anthos is Google’s first step into cloud-native opinionated — and, yes, “hybrid” and “multicloud” — development platforms. Anthos isn’t the first “container-plus-serverless” platform, and we doubt it will be the last. Enterprise customers should add Anthos and Cloud Run to their OpenShift vs. Cloud Foundry bake-off evaluations while leaving room for additional on-prem entrants from Microsoft and Amazon later this year. Expect each platform provider to support the core table stakes: containers, orchestration, service mesh, API management, logging, and tracing. In the near term, look for platform differentiation in developer experiences and DevOps tools, functions as a service, support for events, and integration with unique public cloud services.