September 7, 2017
Most IT systems reside in silos. Businesses face challenges when processing large volumes of heterogeneous data at a rapid pace through various systems. If something goes wrong, fixing the problem can be time consuming and diagnosing the cause can be inconclusive.
When errors occur, IT professionals often leverage a variety of tools to run diagnostics. According to a recent Data Informed article, companies on average own at least 10 separate monitoring tools of which they only really utilize 50 percent.
The issue with running multiple diagnostic tools is the lack of consistency and ability to target root causes. The result is IT teams can take weeks to resolve incidents and conduct a thorough root-cause analysis. The negative impacts are system downtime and overall cost.
To improve efficiency in IT Operations Analytics (ITOA), companies can apply machine learning technologies to help identify issues faster. After an issue is pinpointed, causes and effects can then be discerned.
What Is Machine Learning?
A synonym for artificial intelligence (AI), machine learning helps software predict outcomes based on historical inputs. Supervised machine learning requires actions from a human user to dictate a specific type of outcome. Unsupervised machine learning does not rely on inputs but rather analyzes data to come up with conclusions for complex tasks.
Pinterest uses machine learning to suggest pins or images that cater to a user’s tastes. If a user saves an image of Paris to their pin board, then the software’s algorithms predicts the user wants to see more related images and will automatically generate more pins in that category. The Facebook newsfeed also uses this method based on data from a users’ “likes” or shared posts.
The Missing Link in ITOA
Similar to how social media sites leverage AI to tailor content, IT departments can also use data science to predict system errors and prevent them in the future. Event Correlation Engines are tools commonly used to compare data sets across systems to detect correlations in errors. They can reveal patterns to forecast issues, however, they don’t reveal root causes.
Machine learning technologies can help in IOTA with event correlation and root cause analysis. When data exists in silos, deciphering these correlations isn’t easy. ITOA professionals can create input rules for machine learning to link data separated by environmental silos to predict disruptive events. This kind of machine learning can be either supervised or unsupervised.
The longer machine learning operates, the better it performs. As it learns from previous incidents and how to navigate different data environments, it can determine a probable root cause in a matter of minutes or hours compared to weeks.