Causation and correlation are terms that continually crop up in a variety of contexts when we talk about IT operations management (ITOM). Let’s look at some of the crucial differences between the two.
Causation in its most basic form is when one event causes a second event. It clearly shows that one relies on the other.
Correlation, on the other hand, is when multiple variables behave in a consistent, predictable manner in relation to one another. For instance, whenever one variable increases, the other decreases. A correlation helps you anticipate the change to be expected in the related variable.
The Relationship Between the Two
In both cases, there exists a relationship between the two variables, however they are fundamentally different. While they may exist at the same time, correlation by itself does not imply causation. Likewise, when causation cannot be clearly identified, it is more likely that what you are seeing is a correlation.
Another way to see it is that correlation helps you anticipate the future, because it gives you an indication of what’s going to happen. Causation lets you change the future. These are important concepts in the world of ITOM:
- Knowing the correlation between factors helps you understand how things in the infrastructure behave. An understanding of correlation allows you to make certain decisions - because you can predict how an event will affect another variable. For instance, correlation tells you that if device X goes down, it will impact the availability of network Y or service Z.
- An understanding of causation helps you change how an event will affect other variables in the future. Knowing the cause of a particular issue helps you decide how you want to react.
Don’t Mistake One for the Other
A classic example of the concept of causation in IT ops is in the area of root cause analysis. When something goes wrong in the infrastructure, it is most important that you get to the bottom of what is causing the issue. And the faster you understand the cause, the faster you can affect remediation activities to fix the problem. Knowing the root cause also allows you to fix things once rather than multiple times.
It is essential that root cause analysis is based on causation and not on correlation. Mistaking correlation for causation can result in not fixing the real reason a problem has occurred. In fact, it can cause further unintended consequences resulting in further delays in bringing the infrastructure back to an appropriate level of serviceability.
As a concept, causation therefore plays a key part in the remediation aspects of problem resolution.
Correlation Filters Alarms
Correlation too has its place. It is very useful when it comes to the analytical aspects of problem detection. Since correlation helps us understand how elements are related, you can use this knowledge to rapidly discard redundant pieces of information when debugging an issue.
A good example of this is alarm de-duplication. Because of correlation, a single event can cause multiple alarms – too many alarms to manage. However, knowing the correlation between a data link and the channels that run on it, we can immediately discard all the channel alarms that crop up when a data link goes down.
This reduces the enormity of weeding through tens of thousands of alarms and rapidly brings focus onto the subset of alarms that can then be addressed with further analysis.
Several kinds of correlation come into play when it comes to analysis. It is not uncommon to see event, time, topology and dependency-based correlation used to help triage the onslaught of incoming alarms when things in the infrastructure go awry.
At the end of the day, service availability and uptime are the key performance indicators that are most important when it comes to effectively providing business services and maintaining the underlying infrastructure. Understanding causation, correlation and the differences between them directly impacts your ability to maintain and improve on these KPIs.