The Big Story

Automatic for the cloud

With the possible exception of Docker, no open-source project in recent memory has become as central to a so-called "cloud native" view of the world as Kubernetes. And no vendor has embraced Kubernetes more than Google Cloud, which has made the open-source project created within its walls a key part of its product strategy.

On Wednesday, Google doubled down on that vision, announcing a new Kubernetes service called GKE Autopilot. The idea is to automatically manage the Google Cloud infrastructure for customers that want to use Kubernetes to run their containerized applications, to a degree beyond what is already available from its standard Google Kubernetes Engine product.

Google is trying to overcome some Kubernetes reluctance with GKE Autopilot.The pitch for Kubernetes is that it's the best way to manage demand for containerized applications, a hand-me-down version of what Google uses internally for similar tasks. The knock on Kubernetes, however, has always been its complexity, as a powerful tool that requires Google-like skills to run and operate.

• GKE Autopilot introduces another layer of management into the GKE service, which already helps wannabe Kubernetes users get up and running on Google Cloud.
• The Autopilot service frees GKE users from having to manage the nodes — "a virtual or physical machine" — associated with the pods of containers that all managed Kubernetes services handle on behalf of their users.
• This is another example of the "serverless containers" approach that we talked about in last Thursday's newsletter, where managing the computing resources associated with a containerized cloud service becomes the cloud provider's job. For a fee, of course.

But one thing that remains unclear about Kubernetes is just how many cloud customers actually want to use it.

• According to the Cloud Native Computing Foundation, which manages Kubernetes since Google transferred the project to it back in 2015: "91% of respondents report using Kubernetes, 83% of them in production. This continues a steady increase from 78% last year and 58% in 2018."
• That data, however, is from a survey of CNCF members who by virtue of their membership, have already bought into the CNCF-definition of "cloud native" as centered around containers and Kubernetes.
• But if you've bought into containers and Kubernetes, you're most likely using a managed service from a cloud provider: 90% of Google Cloud customers using Kubernetes are using GKE, while about two-thirds of Kubernetes users on Microsoft Azure and about half of those on AWS pay for the managed version, according to a survey of Datadog customers.
• AWS was the last of the Big Three cloud providers to get on board with Kubernetes, and an increasing number of AWS customers — in fact one-third of container adopters — now prefer to use its Fargate serverless container service, according to the same Datadog report.

Google senses an opportunity here. Those numbers above suggest that while Google (and Microsoft) have supported Kubernetes in part as a way to help companies run applications on multiple cloud providers — in hopes of picking up customers already on AWS — most of their own Kubernetes users are building around a single vendor.

• Managed services, especially ones that automate the deployment of computing resources, tie applications to the managed service providers for a long time.
• One of the reasons it took Parler some time to get up and running after AWS kicked it off its cloud was because Parler's application was written around several high-level AWS services, which (like all cloud vendors) operate in a unique way.
• With GKE Autopilot, Google Cloud is basically telling current and potential customers that believe they need to "do Kubernetes" — whether they actually need it or because the CIO saw a commercial — that it offers a Kubernetes experience that can't be duplicated on other clouds.

This isn't what some Google Cloud execs hoped for, exactly: It's no secret that there's a sizable group of them who pine for a world in which the company never ceded control of Kubernetes to an open-source foundation, believing it could have been a powerful differentiator against AWS and Microsoft.

• Still, that foundation created enormous mindshare for the project and doomed proprietary efforts from Docker and D2iQ (née Mesosphere) in the container-orchestration market.
• And as we've learned from countless examples over the past few years, the true value of an open-source project in the cloud era flows to the companies that can best operate that project at scale, rather than the folks that wrote the code.

Do cloud newcomers want (and need) Kubernetes enough to pick their vendor by the quality of its products based around the project? Google's position is clear.

— Tom Krazit

A MESSAGE FROM INTEL AND MICROSOFT AZURE

For corporate IT managers, there are many motivations to move dynamic workloads to the cloud. It provides an irresistible trifecta of flexibility, scalability, and costs savings for those managing varying workloads. Here's how to keep your data safe while it's in the cloud.

Read more.

This Week On Protocol

Hug it out: After Slack's big outage in January, the last edition of Protocol | Cloud previewed a special project just published this morning: the oral history behind #hugops, enterprise tech's message of empathy during big outages. I talked to 10 of the folks who got a movement started and continue to carry it forward into a new decade.

Mean streets: Rimini Street isn't exactly a household name in enterprise tech, but executives at Oracle and SAP know it all too well. Joe Williams reports on how the company is trying to build a business around providing support for older software, a revenue stream that aging tech companies see as their lifeblood.

Scattered clouds: Microsoft wants to sell customers in specific industries — like retail, finance and manufacturing — packages of cloud enterprise software services designed around their needs. I interviewed Microsoft's Alysa Taylor about these industries and their "need for their digital transformation partners to have deep industry and vertical expertise," as she put it.

Five Questions For...

Amit Walia, CEO, Informatica

What was your first tech job?

My first job was after I graduated with an engineering degree in India and went off to work for Tata Steel, one of the largest manufacturing companies in India, and first-hand learned how to manage teams at a manufacturing setup. It wasn't a tech job, per se, but it taught me important people skills that are extremely critical to success as a leader in any industry.

What changed the most at your company in 2020?

How we have learned to be much more empathetic towards each other, realizing everyone is juggling multiple responsibilities right now on the home front and the work front. Working remotely, we have managed to accomplish many milestones because we had each other's back. Culture is a top priority for me and our company as a whole.

What was the biggest reason for the success of cloud computing over the past decade?

Massive-scale digitalization. In the last decade, the world has seen consumers preferring to shop online rather than walk into a brick-and-mortar, make transactions via their mobile phones and drive cars that know you better than you know yourself. At the heart of this digital-first world is data, and what powers the seamless interactivity of data at scale and at lightning speed is the cloud. Cloud is the ultimate computing machine that you want to ride on for your digital transformation journey, and not a horse buggy.

What will be the biggest challenge for cloud computing over the coming decade?

By 2022, 90% of organizations will utilize multiple cloud service providers and will require significant augmented data management and integration. Such organizations will begin to see huge challenges with data integration and governance as major cloud service providers will try to localize data integration and governance.

Will the pandemic usher in a new era of remote working, or will we all come back together when it is safe to do so?

I think we are surely looking at a hybrid model going forward. There are huge benefits to both face-to-face interactions and working remotely and let's hope when it's safe, we can return to a hybrid model of working.

Around the Cloud

• AWS infrastructure was used to launch parts of the SolarWinds attack, according to several U.S. senators at a hearing investigating the attack. The actual details were unclear.
• Nvidia had another great quarter despite the widespread chip shortages affecting orders around the world, with data-center chips helping to lead the way.
• Chip makers in Taiwan are having to truck in water to their facilities thanks to an extended drought. Lots of water is needed to make high-end semiconductors.
• Chalk one up for D-Wave and its somewhat-controversial approach to quantum computing: The company demonstrated that one of its quantum machines could be used to solve a known mathematical problem several million times faster than conventional computers.
• The Federal Reserve came unstuck this week when the system it uses for interbank transfers and payments went down for several hours Wednesday, prompting a million "bitcoin bro" takes about the efficacy of the system.
• In a far less serious outage, LinkedIn also went down for several hours on Tuesday, an error blamed on a bad configuration change.
• Box CEO Aaron Levie's troubles with activist investors might be entering a new phase after Reuters reported that Starboard is planning a renewed challenge for control of the company barring any big shareholder-friendly changes.
• The Mac OS instances introduced by AWS in December to allow developers to test their applications on the cloud now support Big Sur, the most recent release of the operating system.
• Imagine if you had to stay in prison longer than required because the back-end software used by the corrections department can't figure out new sentencing laws. Well, that's currently happening in Arizona.