The cloud vs. on-premises debate is an old one. It goes back to the days when the cloud was new and people were trying to decide whether to keep workloads in on-premises datacenters or migrate to cloud hosts.
But the Docker revolution has introduced a new dimension to the debate. As more and more organizations adopt containers, they are now asking themselves whether the best place to host containers is on-premises or in the cloud.
As you might imagine, there’s no single answer that fits everyone. In this post, we’ll consider the pros and cons of both cloud and on-premises container deployment and consider which factors can make one option or the other the right choice for your organization.
DevOps, Containers, and the Cloud
Check out Rancher Sandbox – all you need is a GitHub account
First, though, let’s take a quick look at the basic relationship between DevOps, containers, and the cloud. In many ways, the combination of DevOps and containers can be seen as one way—if not the native way—of doing IT in the cloud. After all, containers maximize scalability and flexibility, which are key goals of the DevOps movement—not to mention the primary reasons for many people in migrating to the cloud. Things like virtualization and continuous delivery seem to be perfectly suited to the cloud (or to a cloud-like environment), and it is very possible that if DevOps had originated in the Agile world, it would have developed quite naturally out of the process of adapting IT practices to the cloud.
DevOps and On-Premises
Does that mean, however, that containerization, DevOps, and continuous delivery are somehow unsuited or even alien to on-premises deployment? Not really. On-premises deployment itself has changed; it now has many of the characteristics of the cloud, including a high degree of virtualization, and relative independence from hardware constraints through abstraction.
Today’s on-premises systems generally fit the definition of “private cloud,” and they lend themselves well to the kind of automated development and operations cycle that lies at the heart of DevOps.
In fact, many of the major players in the DevOps/container world, including AWS and Docker, provide strong support for on-premises deployment, and sophisticated container management tools such as Rancher are designed to work seamlessly across the public/private cloud boundary. It is no exaggeration to say that containers are now as native to the on-premises world as they are to the cloud.
Why would you want to deploy containers on-premises?
Perhaps the most obvious reason is the need to directly access and use hardware features, such as storage, or processor-specific operations. If, for example, you are using an array of graphics chips for matrix-intensive computation, you are likely to be tied to local hardware.
Containers, like virtual machines, always require some degree of abstraction, but running containers on-premises reduces the number of layers of abstraction between the application and underlying metal to a minimum. You can go from the container to the underlying OS’s hardware access more or less directly—something which is not practical with VMs on bare metal, or with containers in the public cloud.
In a similar vein, you may also need containers to monitor, control, and manage local devices. This may be an important consideration in an industrial setting, or a research facility, for example. It is, of course, possible to perform monitoring and control functions with more traditional types of software—The combination of containerization and continuous delivery, however, allows you to quickly update and adapt software in response to changes in manufacturing processes or research procedures.
Local Control Over Security
Security may also be a major consideration when it comes to deploying containers on-premises. Since containers access resources from the underlying OS, they have potential security vulnerabilities; in order to make containers secure, it is necessary to take positive steps to add security features to container systems.
Most container-deployment systems have built-in security features. On-premises deployment, however, may be a useful strategy for adding extra layers of security. In addition to the extra security that comes with controlling access to physical facilities, an on-premises container deployment may be able to make use of the built-in security features of the underlying hardware.
Legacy Infrastructure and Cloud Migration
What if you’re not in a position to abandon existing on-premises infrastructure? If a company has a considerable amount of money invested in hardware, or is simply not willing or able to migrate away from a large and complex set of interconnected legacy applications all at once, staying on-premises for the time being may be the most practical (or the most politically prudent) short-to-medium-term choice. By introducing containers (and DevOps practices) on-premises, you can lay out a relatively painless path for gradual migration to the cloud.
Test Locally, Deploy in the Cloud
You may also want to develop and test containerized applications locally, then deploy in the cloud. On-premises development allows you to closely monitor the interaction between your software and the deployment platform, and observe its operation under controlled conditions.
This can make it easier to isolate unanticipated post-deployment problems by comparing the application’s behavior in the cloud with its behavior in a known, controlled environment. It also allows you to deploy and test container-based software in an environment where you can be confident that information about new features or capabilities will not be leaked to your competitors.
Here’s another point to consider when you’re comparing cloud and on-premises container deployment: public and private cloud deployment are not fundamentally incompatible, and in many ways, there is really no sharp line between them.
This is, of course, true for traditional, monolithic applications (which can, for example, also reside on private servers while being accessible to remote users via a cloud-based interface), but with containers, the public/private boundary can be made even more fluid and indistinct when it is appropriate to do so.
You can, for example, deploy an application largely by means of containers in the public cloud, with some functions running on on-premises containers. This gives you granular control over things such as security or local-device access, while at the same time allowing you to take advantage of the flexibility, broad reach, and cost advantages of public-cloud deployment.
The Right Mix for Your Organization
Which type of deployment is better for your company?
In general, startups and small-to-medium-size companies without a strong need to tie in closely to hardware find it easy to move into (or start in) the cloud. Larger (i.e. enterprise-scale) companies and those with a need to manage and control local hardware resources are more likely to prefer on-premises infrastructure. In the case of enterprises, on-premises container deployment may serve as a bridge to full public-cloud deployment, or hybrid private/public deployment.
The bottom line, however, is that the answer to the public cloud vs. on-premises question really depends on the specific needs of your business. No two organizations are alike, and no two software deployments are alike, but whatever your software/IT goals are, and however you plan to achieve them, between on-premises and public-cloud deployment, there’s more than enough flexibility to make that plan work.
Michael Churchman started as a scriptwriter, editor, and producer during the anything-goes early years of the game industry. He spent much of the ‘90s in the high-pressure bundled software industry, where the move from waterfall to faster release was well under way, and near-continuous release cycles and automated deployment were already de facto standards. During that time he developed a semi-automated system for managing localization in over fifteen languages. For the past ten years, he has been involved in the analysis of software development processes and related engineering management issues.