Cell deployment
Unless you are already working on a workload that is multi-Region, you currently have one instance of your workload to develop, deploy and operate. Now with your workload using a cell-based architecture, you have tens, hundreds or even thousands of instances of your workload to deploy and operate, depending on the limits, scale units, and cell size you set. In summary, it is a very complex challenge to deploy in a production environment.
Cell-based architecture brings a new dimension to its context, which is not trivial for
development. If today you have to deploy your source code in an environment (development,
pre-production, and production) in an Availability Zone or Region, now you have to deploy a
cell on all these aspects as well. To avoid problems, it is essential to have an automated
CI/CD pipeline from the beginning. At Amazon, we have a strong culture of continuous delivery,
which you can find out more about in My CI/CD pipeline is my release captain
In the following diagram, we have an Amazon service pipeline type, where each service is deployed in phases until it reaches general availability for all Regions
Amazon service deployment
This is a great way to reduce the impacts of infrastructure failures, bugs, and other errors that can impact customers. With cell-based architecture, these are also deployed in phases, as in the example in the following diagram:
Cell deployment
The benefits of fault isolation and bast radius reduction with cell-based architecture are not only when processing customer traffic, but also when deploying new features and fixing bugs. With your customers or in the partitioning model you chose, your deployment model will also follow this same concept, deploying one or more cells at a time, and when identifying any sign of failure, you can rollback, thus reducing the number of clients that were exposed to this failure.
The previous example was given based on the Amazon deployment model, but the important point here is to deploy in waves, cell by cell or set of cells. It doesn't matter if the cell strategy you chose is non-AZ independency or AZ independency. To delve into other important issues, a good starting point is the Reliability and Operational Excellence best practices of the Well-Architected Framework:
AWS services that can help you with this implementation are:
