A great day to be a developer, but I am left with more questions than answers...
I remember in 2000, the day when IBM announced to the world that it was backing Linux as its strategic operating system into its core systems strategy. A year later IBM invested $1B to back the Linux movement.
IBM in these two actions made Linux a safe choice for everyone, for every developer, system administrator, and for everyone involved in the IT industry we all were empowered to support Linux. Gone were AIX, HP-UX, Solaris, Ultrix and many other downstream proprietary variants of the original UNIX; System V and BSD 4.3.
Google, its announcement of Anthos, and more importantly its strategy around Kubernetes and Istio, have in many ways given me and my team that same feeling of commitment and empowerment. Kubernetes will be the cloud-based operating system and where all significant innovation and contribution will occur for at least the next decade.
In recent years, Google has become the largest contributor to open source and supporter of the Cloud Native Computing Foundation (CNCF). Google has contributed many great enhancements to existing projects, as well as its donations of Kubernetes and Istio. According to stackalytics, Google contributes more than 53% of all code commits. The next closest contributes 7.4%!
Anthos, a new name on last years announcement of Google Cloud Services Platform (CSP) now supports Google’s Kubernetes Service (GKE) on AWS, Azure, and on-premises (through a VMware vSphere VM). Google has announced to the world that hybrid multi-cloud will be the “new normal.” Customers will have initiatives that cross between these environments, and the combination of Kubernetes and Istio will be the platform and service mesh to enable a new era of hybrid cloud applications.
As a developer and early adopter of cloud, I, like my teams, fell in love with AWS. It was the first - and by that definition the best. When the other choice was corporate IT, almost anything was better. AWS liberated developers and democratized innovation and we all went there. We consumed everything AWS would give us. As developers we just wanted to build stuff and let people use it. AWS was everything and more to us.
As AWS grew alone for almost 10 years, Microsoft woke up and turned its guns to AWS. With a bold new CEO, Azure became the overnight favorite of IT organizations and, alongside Office365, became a huge pull for its cloud services.
Then, an evolutionary step happened… Everyone went IaaS and PaaS crazy. Oracle, IBM, Cisco, HP and a host of service providers all went after this market. In April of 2008, Google Cloud was launched. What we saw were the balance sheets of the latest tech companies in the world out-spending each other, with billions of dollars spent in new data centers, new regions, and new innovations. As a consumer of these services, we were spoiled for choice. The big three [AWS, Azure, and Google] have started to put distance between themselves and the others [Cisco, HP, and the service providers] gave up. Now, only Oracle and IBM continue to tell IaaS and PaaS stories.
So, from day one, we had three cloud providers, and the customers started telling us we had to run on their preferred cloud, or better still, we would run on all three! How would we do that? Nothing outside of open source was portable from one cloud to the other, and increasingly the innovation in new technologies out-paced any open source project.
We’ve tried to avoid the question, but it’s kept on coming and now all developers have to consider the landscape for their applications to by hybrid multi cloud.
In 2014, Google open sourced a version of its Borg. The rest, as they say, is history. There are more contributors to Kubernetes than there are on Linux.
Now let's jump to Anthos, a platform for Kubernetes that claims to run on AWS, Azure, Google and on-premises. On the surface, there is a lot to like. However, I and others find ourselves asking more questions than there are answers..specifically
- It’s only GKE, the managed service, and not upstream Kubernetes. So ... is this a lock-in?
- GKE gives no control of the master node(s) in the cluster and therefore no cross-cloud federation support. Most developers need access to the master node.
- You can't upgrade your version of Kubernetes, Google decides when to do that
- If you stop using Anthos, your clusters go away, as they only run on GKE
- Anthos-managed clusters won’t run natively or support EKS or AKS, the Kubernetes managed services from AWS and Azure
- Why have a dependence on vSphere 6.5?
- Anthos costs $10,000 for 100 vCPU’s per month. If you go over you pay them another $10,000, and so on.
- Anthos doesn't scale down either, Kubernetes needs to be for everyone, from single developers to small workgroups, teams, departments, divisions as well as the enterprise. A $10,000 per month entry price pushes it up towards only the largest enterprise organizations.
- Anthos is effectively a cluster configuration management tool, and not an application lifecycle management tool.
- Anthos still misses what everyone needs for truly portable applications, truly portable storage
As a provider of a Kubernetes platform, Cloud data services and storage protocols … what do I think?
- Google and Anthos provide a strong vendor validation of hybrid multi-cloud architectures
- Google has committed support for Kubernetes on other platforms
- Vendors like NetApp must support Anthos as a platform for cluster configuration management
- It further validates the benefits of an upstream Kubernetes offering, and puts the Red Hat and Pivotal strategy of downstream proprietary branches in question
- With Anthos, customers still need Application Lifecycle Management and will have to piece together a number of other incremental solutions, increasing costs and overhead even further
- Customers need choice for their Kubernetes: AKS, EKS, GKE and bare metal
- Kubernetes must be for everyone, not just for the enterprise. Solutions for smaller teams need to exist beyond cobbling together a DIY solution on their own.
- Kubernetes is both powerful and complicated, simpler solutions need to empower more users than the expensive and rare population of Site Reliability Engineers (SRE)
- Developers building stateful applications need really good storage, storage that runs on premise and on the public clouds. Better still, storage that sold and supported by Azure and Google, and available in the marketplace of Amazon. I wonder who does that?
My greatest fear for Kubernetes: the elation that I had for Linux was quickly dashed with the fiasco that was SQL; another standard that never really became a standard. Relational Databases never took SQL to be a portable and open platform.
As a result, databases have ultimately become expensive forms of lock-in. It’s not possible to move from one database to another and Amazon will testify to its publicized move from Oracle. Google, your support for Kubernetes on other clouds and on-premise has cemented its place in history.
We at NetApp are big supporters of Kubernetes and Istio! We will also support Anthos and help customers deliver Kubernetes for everyone, bringing an open application perspective to the complex world of hybrid multi-cloud solutions. We love working with you on enterprise storage protocols on Google Cloud, on application lifecycle management in support of Anthos.