Last updated at Fri, 03 Nov 2017 14:10:04 GMT
DevOps is fundamentally changing.
It is no longer considered a fad by those working with the latest development tools. Now, DevOps emphasizes people and processes as much as the cool tools, and there are many different approaches to bringing them together.
If you Google search for “DevOps flow,” you will find lots of flow diagrams and methodologies. While each one highlights different priorities and names, they all have the same general themes.
- Start with people
- Bring in automation
- Stay lean
- Measure everything
This strategy is called the CALMS approach.
CALMS vs ITSM?
CALMS grew out of the need to bridge the gap between ITSM principles and DevOps. The challenge faced by those familiar with ITSM, who might view as an alternative to DevOps, is understanding where they differ. If the differences are not immediately obvious when viewing the ITSM documentation examples, they will be as you explore the general concepts of the methodology.
The CALMS acronym will help illustrate these different methodologies at a quick glance, and as you observe culture is the first point, it quickly becomes clear that a side-by-side comparison is not possible.
ITSM is generally focused on documentation, change control, and repeatability, where DevOps is focused on people who drive results. With ITSM, repeatability is a main component, but the structure and processes are not primary objectives.
What does CALMS stand for?
Culture is not something you implement; it is the natural environment that houses the entire development team. Entire means all the developers, QA, AND operations. In this eco-system of people and tools, the culture is necessary to remove barriers and facilitate collaboration to hit team objectives.
There is nothing fluffy about culture.
Culture plays a major role in the success of the business so getting it right is important. In the case of DevOps where red tape, time spent on needless activities, and with the silos of responsibilities, bad culture hinders goal completion and company objectives.
Once you have culture in place you can focus on automation. This is not just creating scripts for infrastructure or using Selenium for functional testing.
Automation is the idea that you should program everything.
Find a way to lead all new initiatives with automation and then think about automation constantly to help scale.
As you build automation you also have to remember to be lean.
Running lean means keeping everything to a minimum.
Tools, meetings, and even sprints should all be kept lean. This also is often part of the culture, when you are a minimalist, and every tool, and process you select has a purpose.
It also applies to your teams.
You must keep teams to an effective size to avoid diminishing returns from too many cooks in the kitchen. If the application complexity warrants larger teams, then break them into sub groups, as many large organizations have done.
Measurement is the glue that keeps it all together. When you are a lean organization automating everything, you can lose the ability to remember and document it all. Naysayers of DevOps can quickly point to this challenge as being a reason to avoid DevOps, and they are not wrong.
If a team does not have visibility into everything, something will eventually go horribly wrong.
When measurement is implemented correctly this should not happen. If you measure everything and think in terms of reporting on your development team and processes, you will have the ability to analyze and correlate all processes, tools, and production data together.
However, this cannot be done in the traditional spreadsheet style system. In the minimalist organization, you do not have time to inventory everything, which is prone to human error anyway.
Measurement has to happen using more advanced tools for data collection, like a log analysis platform. One that goes beyond event monitoring and reporting, and can push insights to you. This way you can avoid thinking about pulling information on regular tasks or maintenance, and can delegate the effort to the platform. Once you know what you regularly need to observe, you can build a dashboard that (in a glance) tells you everything you need to know.
With all this data, you need to disseminate what is learned, what went wrong, and how to correct.
The benefit of breaking down the walls between developers and operations is the information flow becomes bi-directional, and you avoid the concept of “throwing things over the fence”. However, this is worthless if information is not shared proactively.
Sharing is not just reporting facts, it is regular exchanging of ideas across teams.
Developers can suggest QA and operations try new tools to improve production quality, and operations can suggest functionality that helps back-end management.
Use CALMS for loose direction, not GPS
All DevOps flows similar to CALMS serve the same general function, which is to be used as a guide, and not a definitive answer.
What is unique with the CALMS description of DevOps is the emphasis on culture.
Culture is not easy, nor straight forward. It takes time to mold unless you have the luxury of hiring all awesome people. Sometimes it is awkward to start, but eventually becomes habit.
Because CALMS is a guide, not an answer, you must remember that it will not be perfect.
I personally believe CALMS is missing an R and an I. CRIALMS, (which sounds like something from World of Warcraft). Results and iterations are also incredibly important. Both, like the above, are culture and directives in parallel.
By nature if you are results driven, you will be analytical, and your results will come faster with leaner operations that collaborate well. In order to execute with the this framework, you need to evaluate your progress on small fast chunks; iterations.
This exercise is important not because CALMS should, or should not, be a standard we throw around like candy, but because when new movements come along (like DevOps) there needs to be a path from the hype to the practice. CALMS helps those who have lived operations and those who do not know what operations even is, but realize they have to work with them.