Do you get excited when you discover a new service from one of the top three public clouds or a new public cloud provider? I do. But every time you feel excited about new cloud offerings, you should also feel a twinge of fear. Because in the tech world, each time we introduce something new we also add a new point of failure for our application and potentially a service we are stuck with. This is why thinking about the long-tail cloud for your organization is important.
What is the Long-Tail Cloud?
You can't predict the future, nor should you try to. What you should do is realize that change in the tech space is constant, and creating a framework for change is critical. Organizations that architect themselves into a corner today, no matter how cutting-edge-tech, are destined to become a dinosaur in a few years. The reason for this is that they did not build their delivery chain and infrastructure in a way that allows them to easily adopt new technologies as they evolve, but rather in a painful waterfall fashion every six years or so.
The long-tail cloud takes into account the long-term cloud strategy of your organization. It's a mindset that embraces change, avoids lock-in, and stewards on-going evolution of their environment.
Creating a Framework for Change
There are several things required in order to create a framework of change. Some tactical and some strategic.
1. Embrace change: Your Ops and Dev teams need to embrace the idea that there is a good chance the way they are currently working and the infrastructure they are using will likely change dramatically in the next six months to a year. Just compare how quickly docker containers went to Kubernetes, and now Kubernetes to serverless. In order to compete, development teams should be able to leverage new technologies when it makes sense.
2. Know what you have, and what you don't: Having visibility into the services you are using — and that you are not using — is very important. This helps you quickly identify gaps and overlap. Visibility ideally also means you see what you are missing, and can compare your services with others. Visibility can come in the form of actual UI/UX where you can eyeball services that you have, and identify new ones. But it should also have built-in interfaces for your developers to do the same in their standard dev environments like CLI. And finally, it should help you get an understanding of your costing.
3. Right tool for the job: This is a biggy. Because your application runs 90% on AWS, Google, or Azure exclusively, does not mean that if the right purpose-built cloud service comes along you should avoid it just because it's not running on your public cloud of choice. Conversely, organizations also make the mistake of picking less than ideal public cloud services just because they're available in their public cloud of choice, versus focusing on what is best for their users.
4. Specialization is key to giving your users the functionality they want: As we move beyond the generalist cloud services, in many modern applications their highly specialized functionality will require highly specialized cloud services. Specialized IoT, Machine-learning, BigData, etc. services, and APIs will become more commonplace. Organizations need to easily adopt these tools for their application to stay ahead.
The Risk of Cloud Exclusivity
There are some risks that come with being architected for a single public cloud.
1. It impacts your application architecture and features. Many dev teams find they are tailoring functionality to fit their public cloud when it should actually be the other way around. The cloud provider should be able to support anything you throw at it.
2. Your cloud provider could make drastic shifts in their strategies, which could be inconsistent or even contrary to your organization's. They also hold the keys to your pocketbook, and without options, they can modify pricing with little recourse for you.
3. Your security and stability exposure is equivalent and never better than the single cloud you run on. With only a single cloud, your application availability is only as good as theirs, and you are powerless to any exploits they face, or outages. We have learned that whole public cloud regions can go down.
These three reasons are why multi-cloud has become such a popular term, and why organizations are really considering it. But, they also need to think about the mechanism for supporting it.
Engineering Change
There is a sense of security that comes with committing to a single public cloud. It makes it easy to adopt new technologies IF the public cloud provider has them, and it avoids some internal communication because the broader cloud offering has already been approved.
But this also comes at a cost. It can force application teams to architect their applications in a way that is not optimal for its users. It can also force a lock-in that prevents organizations from embracing change should it make sense, or they need to.
Organizations can engineer change by changing their mindset, deciding not to be cloud exclusive, and considering tooling that gives them the visibility and ability to adopt cloud services that are the right fit for their application and organization.
