4 Things Machine-Learning Algorithms Can Do for DevOps Right Now
Originally published by Ronny Lehmann at www.loomsystems.com on March 16, 2017.
AI and Machine Learning Algorithms are providing the foundations for DevOps fruition.
As the IT industry struggles to elevate performance, companies are looking to DevOps to deliver on the promise of a newly efficient process (frequent release cycles to feed the higher demanding consumers) . But speeding things up is far easier said than done.
Transforming to digital and transforming within a digital environment are both difficult evolutions for enterprises; fruition will require a new mindset — but one that is facilitated by technological progress.
I was recently quoted in an article for Computer Weekly, commenting that despite the relatively slow adoption curve we witnessed in the field of machine-learning, DevOps practitioners have a lot to gain by embracing even the most basic of capabilities that AI has to offer. In fact, whether or not to embrace AI is no longer the question but rather where to start — right now.
Here are some basic processes that I strongly believe DevOps will be relying on machines to perform in the recent future:
1. Maintaining an Alerting Rule-engine
An ongoing and repetitive task (aka mundane), which almost always goes out of control when done manually. Maintaining the alerting “rule-engine” is a classic fit for the smart machine — to do it right, one needs to take into consideration inter-dependencies, baseline behavior; and to do all this continuously. Machines can also dig deeper, identify changes in trends or bend-before-break situations that aren’t feasible to detect with today’s arsenal.
2. Prioritizing Alerts
Manually prioritizing alerts in today’s flooded environment is like rearranging the deck chairs on the Titanic. The capacity that smart machines have to apply parameters to prioritization is virtually endless and dynamic. Alerts are analyzed for significance in real time and machines will correlate the alert to everything the application or component is related to and affected by. This level of sophistication enables DevOps to use alerts proactively and precisely.
3. Analyzing the Root Cause
The importance of finding the root cause of a problem cannot be stressed enough. From our observations, we’ve learned the significance of clearing “the fog of war” — which is the countless of alerts (and complaints) one gets when something breaks. Machines can help clear the fog by drawing maps of correlations and causality between the various inputs. Machine learning enables DevOps to cut through the noise and determine the sources and correlation of all alert.
4.Introducing Recommendation Algorithms
Although there are, in fact, so many tasks and processes that can be smartly automated to enable DevOps to do their job, recommendation algorithms is definitely one of the magic makers. Smart machines will go beyond mundane task excellence to provide ANSWERS. Receiving alerts that are relevant, correlated and prioritized is one thing, but having them delivered alongside the relevant resolution is a whole new level of efficient for DevOps.
Digitization, DevOps and the likes are here and now with crucial issues that need to be solved NOW if we are to move forward. It’s exciting times for AI and machine learning and we’re honored to be among the drivers of this evolution.
Interested in seeing a recommendation algorithm at work? Take a look at Loom Systems, the next generation log analysis & monitoring platform, leveraging AI to predict & solve problems in all of your applications before your end-users even know about them. Schedule Your Live Demo!