GPUs on GKE (an acronym Google used to be quite fond of, but seems to be deemphasizing now) have been available in closed alpha for more than half a year. Now, however, this service is in beta and open to all developers who want to run machine learning applications or other workloads that could benefit from a GPU. As Google notes, the service offers access to both the Tesla P100 and K80 GPUs that are currently available on the Google Cloud Platform.
The advantages of the kontainer/GPU combo is that you can easily scale your workloads up and down as needed. Most GPU workloads probably aren’t all that spikey, but in case yours are, this looks like a good option to evaluate.
Overall, the Kubernetes Engine (which seems to be Google’s prefered nom du guerre for the GKE now) saw its core-hours grow 9x year over year in 2017. That’s no surprise, given the hype around containers and the fact that the GKE, as it was called at the time, only launched in 2016, but it does show that Google may just have a winner here.
Red Hat, a company best known for its enterprise Linux products, has been making a big play for Kubernetes and containerization in recent years with its OpenShift Kubernetes product. Today the company decided to expand on that by acquiring CoreOS, a container management startup, for $250 million.
If the next generation of software is going to be in a hybrid cloud world where part lives on prem in the data center and part in the public cloud, having a cloud native fabric to deliver applications in a single way is going to be critical. Red Hat’s president of products and technologies, Paul Cormier said that the combined companies are providing a powerful way to span environments.
“The next era of technology is being driven by container-based applications that span multi- and hybrid cloud environments, including physical, virtual, private cloud and public cloud platforms. Kubernetes, containers and Linux are at the heart of this transformation, and like Red Hat, CoreOS has been a leader in both the upstream open source communities that are fueling these innovations and its work to bring enterprise-grade Kubernetes to customers,” Cormier said in a statement.
As CoreOS CEO Alex Polvi told me in an interview last year, “As a company we helped create the whole container category alongside Google, Docker and Red Hat. We helped create a whole new category of infrastructure,” he said.
His company was early to the game by developing an enterprise Kubernetes product, and he was able to capitalize on that. “We called Kubernetes super-duper early and helped enterprises like Ticketmaster and Starbucks adopt Kubernetes,” he said.
He explained that Tectonic included four main categories include governance, monitoring tools, chargeback accounting and one-click upgrades.
Red Hat CEO Jim Whitehurst told us in an interview last year that his company also came early to containers and Kubernetes. He said the company recognized containers included an operating system kernel, that was usually Linux. One thing they understood was Linux and so they started delving into Kubernetes and containerization and built OpenShift.
CoreOS raised $50 million since its inception in 2013. Investors include GV (formerly Google Ventures) and Kleiner Perkins, who appear to have gotten a nice return here. The most recent round was a $28 million Series B in May 2016 led by GV. One interesting aside is that Google, which has been a big contributor to Kubernetes itself and whose venture arm helped finance CoreOS, was scooped by Red Hat in this deal.
The deal is expected to close this month, and given we only have one day left, chances are it’s done.
Heptio holds a special place in the Kubernetes startup ecosystem. Its co-founders, Craig McLuckie and Joe Beda, are, after all, also two of the co-founders of the Kubernetes project (together with Brendan Burns), which launched inside of Google. Heptio also raised $8.5 million when it launched in 2016 (and another $25 million last year), but it was never quite clear what the company’s actual business plan looked like beyond offering training and professional services. That’s becoming quite a bit clearer now, though, as the company today announced the launch of the Heptio Kubernetes Subscription.
I always assumed that Heptio would launch some kind of Kubernetes distribution in the near future — and that’s kind of what this is, but the company is also putting a different spin on this. Indeed, Heptio CEO McLuckie described the subscription service as an “un-distribution” when I talked to him earlier this week. The idea behind this concept is that enterprises do not want to lock themselves into a distribution — or a cloud, for that matter. Kubernetes and containers are what allows them to build multi-cloud solutions that avoid lock-in, but a distribution tends to make them dependent on a single vendor.
Enterprises are also going through an interesting cognitive shift, says McLuckie, because just like Google realized that its container technology would get better from open sourcing it, enterprises are now realizing that if they hire thousands of developers, it makes sense for them to be more connected to the technology that they are betting their businesses on — and the best way to do that is to be active in open source. That’s why Intuit, for example, recently acquired Applatix, the company behind the Argo workflow engine for Kubernetes.
McLuckie argues that this is creating a bit of tension because distributions tend to be about taking open-source technologies and making them enterprise-ready (and maybe adding some proprietary code on top of that) — because the community typically wouldn’t be able to do that. But now Google, Microsoft, Amazon and other cloud providers are putting in this work, so a distribution vendor becomes less of a necessity.
Add on top of that the fact that it usually takes some distribution vendors quite a while to offer the latest versions of a tool like Kubernetes (they have to integrate and test updates with their own proprietary code, after all) and the fact that you’re not just getting Kubernetes with the distribution but also a predefined set of other software for networking etc., and many enterprises would rather opt for a more flexible solution that also allows them to run their code on any platform.
And that’s where Heptio’s new offering comes it. It’s based on the company’s reference architecture and solely uses open-source tools (including Heptio’s own Sonobuoy, for example) and the latest version of Kubernetes, and that allows for deployments on any major cloud provider or on-premises. The company built a framework that handles the installation and then offers 24×7 support to its subscribers. Ideally, this means you get the best of a distribution (that is, a tested, enterprise-ready Kubernetes install, plus support) without any of the downsides. “Companies want a simple, modular installation framework,” McLuckie told me — and they want to be able to shift their consumption model as necessary.
With this, the company is focusing on organizations that have already decided on Kubernetes. Almost by default, these tend to be pretty sophisticated companies that want a multi-cloud solution based on Kubernetes, but they often need a bit of help in getting started with a product as complex as this.
Heptio also promises to bring its own improvements to Kubernetes and the installation procedure to the community. Over time, that’ll mean that one of its major value propositions will go away, but McLuckie is quite aware of this. That’s why Heptio is already working on its next generation of products and services that bring additional value to its users. What these will look like, though, remains to be seen, but McLuckie promised that he’ll have more to share by the middle of the year.
As for pricing, McLuckie tells me that he doesn’t want to charge a per-node fee. “We sell it in ‘t-shirt sizes’ where the scaling factor is the size of the environment and the number of unique configurations — different hosting environments, network configurations, etc.,” he said. “Our goal is to simplify things for the enterprise buyer and make sure that we are delivering honest value. We do not believe, for example, that it is intrinsically more complicated to run a 25 node cluster production cluster than it is to run a 100 node cluster and our pricing reflects that reality.”
Mirantis co-founder Adrian Ionel left his CEO role in October 2015, at a time when the company was still solely focused on OpenStack. Now, he is coming back to reprise his role as the company’s CEO at a time when Mirantis is looking beyond OpenStack and toward the still nascent cloud native ecosystem around Kubernetes for its next big business opportunity. Mirantis’ current CEO Alex Freedland will remain a board member.
Like other companies that made an early bet on OpenStack, Mirantis, which once described itself as “the pure play OpenStack company,” went through a series of ups and downs (it now also describes itself as the “#1 pure play open cloud company”). But just like OpenStack has now found its place, Mirantis weathered these storms as an independent company and is now able to take what it learned in the last few years in delivering OpenStack to a wider swath of products.
“The company needs to find its path forward, leveraging the present — the enormous assets we have in the cloud infrastructure business — and articulate a compelling and exciting and engaging future vision that brings us into the future” Ionel told me. “That’s quite similar to where we were in 2010.”
For now, Mirantis’ foundation will remain its private cloud business, where it will continue to offer its managed cloud solution for OpenStack. “Private cloud is still a very large market and we have a very compelling answer,” Ionel said. And with companies like AT&T (which runs a 10,000 node OpenStack cluster with the help of Mirantis), the Shenzhen Stock Exchange, VW and others on its customer roster, this will likely remain a good part of the company’s business in the near term.
But in building out its continuous delivery platform for OpenStack, Mirantis also built the foundation for the future of the company. “We believe that our open source toolbox — our unique approach to open source and the open source technology out there — are central to helping companies build and run applications on any cloud and that we will use the same playbook – the continuous delivery playbook and the open approach that made us successful in OpenStack to help companies run apps in a fully automated fashion in any cloud — public or private,” Ionel explained.
Mirantis already started this journey under Freedland. With DriveTrain and its Container-as-a-Service offering, it built the kernel of the continuous delivery platform that Ionel envisions and that Mirantis needs to focus on as its customers start asking for multi-cloud solutions.
Ionel also seemed very excited about the fact that this move brings Mirantis closer to the actual applications and toward application delivery in general. “We want to give customers and autopilot for applications they want to deploy,” he said. “We will invest heavily in cloud native continuous deliver that enables developers to deploy their code with full automation and ease.”
Returning to Mirantis wasn’t an obvious choice for Ionel and he admitted that it took him a few months to make a decision. In part, that’s because he built another company, Dorsal, after leaving Mirantis. He also noted that he prefers to look forward in life, “so if you rejoin something, you want to think very carefully why you are doing it.” In the end, though, he decided that he would feel very much at home at Mirantis and that he would be able to help the company in this next phase.
For a technology that the average person has probably never heard of, Kubernetes surged in popularity in 2017 with a particular group of IT pros who are working with container technology. Kubernetes is the orchestration engine that underlies how operations staff deploy and manage containers at scale. (For the low-down on containers, check out this article.)
In plain English, that means that as the number of containers grows then you need a tool to help launch and track them all. And because the idea of containers — and the so-called “microservices” model it enables — is to break down a complex monolithic app into much smaller and more manageable pieces, the number of containers tends to increase over time. Kubernetes has become the de facto standard tool for that job.
As Kubernetes has gained momentum, it has become a platform for innovation and business ideas (as tends to happen with popular open source projects). Once you get beyond the early adopters, companies start to see opportunities to help customers who want to move to the new technology, but lack internal expertise. Companies can create commercial opportunities by hiding some of the underlying complexity associated with using a tool like this.
We are starting to see this in a big way with Kubernetes as companies begin to build products based on the open source that deliver a more a packaged approach that makes it easier to use and implement without having to learn all of the tool’s nuances.
To give you a sense of how quickly usage had increased, 451 Research did a container survey in 2015 and found just 10 percent of respondents were using some sort of container orchestration tool, whether Kubernetes or a competitor. Just two years later in a follow-up survey, 451 found that 71% of respondents were using Kubernetes to manage their containers.
Google’s Sam Ramji, who is VP of product management at Google (and was formerly CEO at Cloud Foundry Foundation), says it feels like an overnight sensation, but like many things it was a long time in the making. The direct antecedent of Kubernetes is a Google project called Borg. Ramji points out that Google was running containers in production for a decade before the company released Kubernetes as an open source project in 2014.
“There was almost a decade of container management at scale in Google. It wasn’t an experiment. It was code that ran the Google business at scale on Borg. Kubernetes is built from scratch based on those lessons,” Ramji said.
Cloud native computing
One of the big drivers behind using Kubernetes and cloud native tools in general is that companies are increasingly operating in a hybrid world where some of their resources are in the cloud and some on-prem in a data center. Tools like Kubernetes provide a framework for managing applications wherever they happen to live in a consistent way.
That consistency is one big reason for its popularity. If IT was forced to manage applications in two different places using two different tools (or sets of tools), it would (and does) create a confusing mess that makes it difficult to understand just what resources they are using and where the data is living at any particular moment.
One reason the Cloud Native Computing Foundation is called that (instead of the Kubernetes foundation), is that Google and other governing members recognize that Kubernetes is only part of the cloud native story. It may be a big part, but they want to encourage a much richer system of tools. By naming it more broadly, they are encouraging the open source community to build tools to expand the ability to manage infrastructure in a cloud native fashion.
Big companies on board
If you look at the top 10 contributors to the project, it involves some major technology players, some of whom cross over into OpenStack, Linux and other open source projects.These include Google, Red Hat, CoreOS, FathomDB, ZTE Corporation, Huawei, IBM, Microsoft, Fujitsu, and Mirantis.
Dan Kohn, the CNCF’s executive director, says these companies have recognized that it’s easier to cooperate around the base technology and compete on higher level tools. “I would draw an analogy back to Linux. People describe Kubernetes as the ‘Linux of the cloud’. It’s not that all of these companies have decided to hold hands or are not competing for the same customers. But they have recognized that trying to compete in container orchestration doesn’t have a lot of value,” he said.
And many of these companies have been scooping up Kubernetes, container or cloud-native related companies over the last 12-18 months.
container development team workspaces
operate and deploy cloud native apps at scale
workflow tool for Kubernetes
cloud-like continuous updating
mutli-cloud applications management
support and tooling for Kubernetes
All of this adds up to a set of businesses being built around a tool that didn’t even reach 1.0 until July 2015 (although there were several 0.x releases prior to that). Since then, use has steadily climbed.
Earlier this year, the CNCF announced that 36 companies agreed to a Kubernetes certification standard — when was the last time 36 tech companies agreed to anything? They did this to prevent any individual member from creating a non-compatible or inconsistent version that would either behave differently than expected or would not be portable from one version to another. This is typically known as forking and the organization, recognizing the growing popularity of Kubernetes, wanted to ensure to the extent possible that, that didn’t happen.
Beyond finding ways to commercialize the core open source version of Kubernetes, there are a range of other tools being developed from host management and secure images to logging and monitoring to name but a few.
All of this points to a rich of set of tools being developed around an open source project that is barely two years old. This is what happens when you create an open system. Innovation tends to happen as people need tools applications for running that. We have seen it with Linux. We’ve seen it with Hadoop and OpenStack, and we are seeing it with Kubernetes — and this year it took off in a big way.
Featured Image: Stan Olszewski/SOSKIphoto/Flickr UNDER A Copyright LICENSE
It is the latest in a long line of big name companies, joining the likes of AWS, Oracle, Microsoft, VMware and Pivotal, all of whom joined in a flurry of activity earlier this year. Most of these other companies have more of a cloud infrastructure angle. Salesforce is a SaaS vendor, but it too is seeing what so many others are seeing: containerization provides a way to more tightly control the development process. Kubernetes and cloud native computing in general are a big part of that, and Salesforce wants a piece of the action.
Salesforce’s Mark Interrante wrote in a blog post on Medium announcing the partnership that it’s always a good sign when developers adopt new technologies like Kubernetes quickly. It proves that they see how these tools help create products faster and easier, certainly a goal for a company like Salesforce that is generating new products at a rapid pace.
In fact, Interrante wrote that Salesforce’s development teams have adopted many of the CNCF tools including Kubernetes. “We’ve seen how containerization simplifies the orchestration of software across a large fleet of servers. Kubernetes makes a great foundation for continuous innovation/continuous delivery which then improves our software delivery,” Interrante wrote.
Salesforce, much like the other large software companies that have joined the CNCF, wants to benefit from being a part of the process and helping to influence the development of key tools moving forward. “This kind of collaboration, with Salesforce as an active participant in open technology ecosystems, is key to helping us move forward,” he wrote.
The CNCF and Kubernetes have become a force in the containerization space in a relatively short time. While Docker, another popular container technology helps developers create containerized software, Kubernetes helps deploy and manage it all and Salesforce wants a closer relationship with the organization that’s running the project.
With the rise of Kubernetes as the de facto standard for container orchestration, it’s no surprise that there’s now a whole ecosystem of companies springing up around this open source project. Heptio is one of the most interesting ones, in no small part due to the fact that it was founded by Kubernetes co-founders Joe Beda and Craig McLuckie. Today, Heptio announced that it is teaming up with Microsoft on its Heptio Ark project, which it launched earlier this year.
Heptio Ark is a utility for managing backups and disaster recovery that helps you bring your Kubernetes clusters and volumes back up after your run into a major issue in your data center.
The plan is for Microsoft and Heptio to work together on strengthening Ark’s core capabilities, but also on making it a tool for moving Kubernetes applications across on-premise environments and — unsurprisingly — Microsoft Azure and the Azure Container Service (AKS — because Microsoft hasn’t come around to renaming it to ‘Azure Kubernetes Service’ just yet).
“Few real-world companies live solely in the public cloud,” said Heptio CEO Craig McLuckie. “It is incredibly important that the tools and practices they adopt when selecting their public cloud services work on-premises as well. Microsoft’s commitment to working with the open source community will not only benefit Azure customers, but strengthens the Kubernetes community.”
What’s also interesting about this move is that it brings the three Kubernetes co-founders together, with Beda and McLuckie at Heptio, and Microsoft’s Brendan Burns, who worked with the Heptio founders at Google when they launched the Kubernetes project into open source (and who was the lead engineer for Kubernetes at Google).
“I’m excited to see Heptio and Microsoft deliver a compelling solution that satisfies an important and unmet need in the Kubernetes ecosystem,” said Burns. “We’re working with Heptio to ensure that the integration of Ark and Azure is a best-of-breed solution for backing up on-premise Kubernetes clusters into the cloud.”