Open Source Summit Europe + ELC Europe 2017

Due to the generalization of the cloud, some people today try to use multi-region and multi-cloud environment more and more.
These systems have three merits:
1. to improve availability
2. to avoid cloud lock in
3. to optimize the cost
That is, even when you have some problems with one cloud, you can use another cloud as a substitution.
In addition, you can change from one to another easily, so you don’t need to be afraid of lock-in.
Finally, using multi-region and multi-cloud helps you optimize the whole cost.
For example, AWS and GCP have various instance sizes and types such as Spot instance and Preemptible VM, and prices are different for each of sizes or types.
According to the change in the circumstances, you can choose and allocate such instances properly using multi-region and multi-cloud systems.

In spite of these potential advantages, you don’t know how to use effectively multi-region and multi-cloud yet.
There are two main reasons in this:
1. It is too complicated to define infrastructure across several cloud providers.
2. It is too difficult to scale infrastructure if you want to choose reasonable providers and properties.
Terraform can solve one of the problem.
Terraform is a useful tool for handling the former problem and help your deployment.
However, management of infrastructure after deployment is beyond their scope.
I would like to propose a new multi-cloud operation using Infrakit, Docker’s software.
Infrakit will solve both problems all at once.
In other words, you can freely add custom rules for scaling and modify infrastructure without being disturbed by stopping service.
First, I will talk about how to define and deploy infrastructure with Infrakit.
Then I will describe the implementation of infrastructure management rules and the way to add custom rules.
Finally, I will demonstrate that you can deploy and scale infrastructure across cloud providers and on-premice environment with Infrakit.