sumetho - Fotolia
Expect more container evolution, growth in 2019
Containerization shows no sign of winding down. With technologies, product options and providers focused on technological improvement, container ecosystems will develop rapidly in the coming year.
Containers grew in popularity and rose as a viable go-to strategy for virtualization for enterprises in 2018. Can there be anything left to improve in 2019? Hang onto your hat, because container evolution hasn't slowed yet.
Expect the open source community and vendors to formalize the container ecosystem in the near future. The focus of containers throughout the technology's rise has been low overhead and easy virtualization. Containers fit into an application ecosystem inherited from data centers, VMs and the cloud. In 2019, containers will become independent from this legacy. Containers will still support application deployment as a hosting option, but now with a complete set of tools to build and run distributed applications.
Orchestration -- the layer of the container technology stack where containers are deployed, duplicated and redeployed if they need to scale up or replace a failed container -- is the great unifier of container technology, and it will begin to harmonize on two pathways in 2019. Kubernetes, which is a deployment-centric mechanism for orchestration, is the leading technology choice, but Apache Mesos and Mesosphere's DC/OS and Marathon offerings provide a virtual-infrastructure-centric alternative. By virtualizing all deployment options, Mesos reaches beyond containers to support any kind of application hosting.
Container evolution includes nascent tools -- such as KubeVirt and Virtlet, which is a feature of the Mirantis Cloud Platform -- that run VMs under Kubernetes. In addition to these third-party options, expect to see changes in Kubernetes itself to provide something more like the virtual-infrastructure approach that Mesos takes. This technological growth represents a shift from the Kubernetes approach to focus only on deployment and redeployment on open resources. A virtual-infrastructure-enabling model of Kubernetes must be open to extension. For most users, container adoption will be first and foremost built around Kubernetes.
With an abstraction of infrastructure, which is the direction in which Kubernetes is heading, development, monitoring, networking and every other piece of the operations for applications must consume the abstraction and not the real infrastructure. For this approach to work, containers must look, for all intents and purposes, like real computers, at least to the applications that they host. To reach this state, the IT industry needs to see a general opening of the container file system to console or user interaction, and support for virtual desktops on containers. These changes enable what could be called an empty container, wherein containers are virtual hosts in every respect.
Virtual infrastructure that enables containers to act as virtual hosts should enable abstraction so that multiple hosting resources and application components look like a single virtual host and application.
Two additions for the virtual host
Serverless computing, while it does actually take place on physical infrastructure, is so-called because it assigns virtual resources to application components on-demand, and steers resources to the components that need them dynamically. We are just beginning to see a new dimension in orchestration for serverless computing: workflow orchestration, the first orchestration addition needed for container evolution in 2019.
Workflow orchestration means Kubernetes must be aware of how computing work is distributed among available resources. Kubernetes must recognize when container deployment and scaling up or down makes the workflow change, versus when the workflow drives deployment of computing resources. In serverless, the workflow dictates what resources are consumed when, and by which application components. Containers and serverless are deployment models with overlapping use cases, and workflow orchestration capabilities must seamlessly shift between the two approaches.
Another orchestration improvement that will enable a new wave of container evolution is improved event-handling. Events are supported in Kubernetes, and many monitoring extensions for Kubernetes can recognize infrastructure and application conditions and generate events that drive a response. In addition, tools like Brigade, from Microsoft, create scriptable Kubernetes clusters. And event-processing software is available, such as Istio service mesh, to provide a workflow fabric, as well as other methods to enable microservices and functions-based computing.
All event-based management and automation technology works with Kubernetes, but there are too many options and too many hassles around integration for many users. Logically, event recognition should become formally tied to orchestration. If it comes to fruition in 2019, expect better monitoring and application lifecycle management as a result.
Networking and the cloud
Virtual container networking is also due for significant changes. Kubernetes and other orchestration technologies for containers have long supported various virtual network plugins, which have significant differences in capabilities. Plugin variability can create major problems when containers are used with cloud infrastructure.
Hybrid cloud is another powerful driver of change, paired with containerization as well as separate from it. Hybrid cloud proponents will promote a virtual network model in 2019 that explicitly deals with cloud to data center connectivity, connections to virtual private networks, and software-defined wide-area network integration with containers.
Container adoption will increase in 2019 alongside all these technological factors that promise major, rapid changes. Options and choices around container technology could require considerable training time, and integration work, at enterprise IT shops.
Expect a shake-out in container product options and providers. A container ecosystem is a lot more than Docker or even Kubernetes, and vendors including VMware, Hewlett Packard Enterprise, IBM and Red Hat all are collecting necessary pieces of the evolving container ecosystem and packaging them to appeal to customers. These packages will increasingly dominate the market as enterprises deploy significant workloads on containers.