DevOps: Why Silos Suck And How To Break Them

Creative Commons License eirikref

Divide and conquer, Caesar’s strategy to break huge problems down into smaller parts, is an outdated model for structuring teams and organizations. Breaking teams apart by area like development, QA, operations, product management, etc, creates silo like divisions of labor. Unfortunately, these divisions create so many “walls of confusion” between the silos that your speed and agility is seriously hampered.

But, specialization, divide and conquer for a field of knowledge, is still necessary to cover a field well enough. Consider security or networking – you’d really need to dive very deeply into them to understand all the tiny, little details which make or break your infrastructure.

DevOps To The Rescue

Having specialists, but no silos, is possible using the right culture, management approach and tools. DevOps is a common name for exactly that.

The DevOps Culture

On a cultural level, DevOps means that everyone has to own the overall business goals. It’s of no use to say: “well, my part works, but they can’t get it running”. Instead, everyone has to do his best to realize business value.

The DevOps Process

On a process level, DevOps needs developers to work closely together with QA and operations. They have to make sure that test scenarios are written alongside the code (and not weeks after finishing it), and they have to make sure that they know how their product will behave in operations. The same is true for the other parties as well. QA needs to make sure to be in the process right from the beginning and operations should have critical influence on the runtime architecture of the product.

Tools Facilitate DevOps

If people start owning the whole process instead of just their area of expertise, you’re already well ahead of most of your competitors. But DevOps is more than the process and the mind set: It can be strengthened a lot by using the right tools. Programmable infrastructure is the key here. APIs and automated tools like Puppet and Chef to setup servers, one-click deployments (e.g. using capistrano), all help facilitate the cooperation of the various experts.

Have you already introduced ideas or tools to help you bridge the gap between development and operations? What are your experiences with DevOps? Let us know in the comments!

3 thoughts on “DevOps: Why Silos Suck And How To Break Them

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.