Do you work in an IT company and suffer from bad processes and bad management? This article may give you some hints how to proceed.
At 8. and 9. April 2019 I joined the "DevOps" training in Cologne.
We talked about:
A lot of stuff is written in a book which we worked out.
Here I give a short summary of the content of this training session:
We found out, that we are facing a lot of new problems in our companies during the last years. And our solutions don't help. The complexity of our IT-departments has risen enormously. And the solution was always: "Let's find some rules, how to deal with it." But this enhances the complexity.
For these kind of problems, it is not a solution to:
"Invest more into infrastructure", "Find a replacement for the employees", "hire more developers", "accelerate the work"…
Companies have the problems, that:
• Ideas are created by a management, that has no clue of developing
• A lot of information gets lost from top to bottom
• Prioritisation is decided by the management and not the people, that know the pain
• Code is not or bad merged
• Tests are done too seldom
• It takes too much time to test
• Deployment is too complex
• Deployments often fail and each team blames the other
• Time to market is huge
• There is no flexibility on production environment
• Finished releases and features cannot be changed ad hoc
• Testing by the customer is not possible
We have been shown the SAFe DevOps Health Radar, a tool that shows on a onepager, what kind of problems exist and what kind of actions can be done to mitigate the threads/problems.
SAFe DevOps Health Radar
There are four big categories of topics: Continuous Exploration, Continuous Integration, Continuous Deployment and Release on Demand. For each of these categories, there are four further subcategories.
In our training group we created an example for a cycle of bringing in a new application to Single Sign On (SSO). We talked about each sub-category, where we are, with which grade from 1 to 5, with 1= bad and 5=good (green dots).
Then we created a canvas with current and future steps:
DevOps Transformation Canvas
The current steps in the lower left corner describe, how an application is attached to SSO.
The steps are written on yellow paper.
The responsibilities are written on pink paper.
The Lead Time (LT) is the latency time, that it needs to wait, until a task is started.
The Process Time (PT) is the time the tasks needs to be done. Both are written on blue paper
The Percent for Complete and Accurate (C&A) is the percent of work, that the next step can use (100% - failure).
With this model we can see how long it takes and how many output we have all in all. (In this example 45 days with 30% output (for uploading some simple roles)).
In the right upper corner, we should have done the same with the future process after executing improvement actions. But we didn't understand it in our training team, so we just wrote down the improvement actions.
With the DevOps Health Radar we have been given 16 categories were we are able to pull the lever. For each category, there are several improvement actions. All in sum it is about 52 transformation actions, that we can use (and have to train for the exam ;-)). Here I show the topics of it:
For each of these action items we should think if we can use it. We should guess the cost of delay (What does it cost, if we don't do it) and the job size (What does it cost to do it). Then we divide the cost of delay by the job size and get the Weighted Shortest Job First (WSJF), a prioritized backlog item, a way to get from the situation where we are now into a bright future ;-)
Achim Mertens
