Follow Gregory Piatetsky, No. 1 on LinkedIn Top Voices in Data Science & Analytics

KDnuggets Home » News » 2018 » Feb » Tutorials, Overviews » Applying Machine Learning to DevOps ( 18:n09 )

Applying Machine Learning to DevOps


This article explains the synergy between DevOps and Machine Learning and their applications like tracking application delivery, troubleshooting and triage analytics, preventing production failures, etc.



By Prasanthi Korada, Andhra University

The world of using static tooling for packaging, provisioning, deployments, and monitoring, APM and log management will be over. With Docker adoption, the Cloud and API driven approaches and micro-services to deploying applications at a large scale, ensuring high reliability, requires an excellent take. So, it’s essential to include creative managing tools for cloud instead of reinventing the wheel every time. With the rise of ML and AI, more DevOps tooling vendors are incorporating intelligence with their offerings for further simplifying the task of engineers.

Header image

The synergy between Machine Learning (ML) and DevOps is potent, and their related capabilities include:

  • IT Operations Analytics (ITOA)
  • Predictive Analytics (PA)
  • Artificial Intelligence (AI)
  • Algorithmic IT Operations (AIOps)

Machine Learning is the practical application of Artificial Intelligence (AI) in the form of a set of programs or algorithms. The aspect of learning relies on training time and data. ML conceptually represents acceleration and codification of “Culture of Continuous Learning” by Gene Kim’s. The teams can mine linear patterns, massive complex datasets, and antipatterns, and refine queries, uncover new insights, and continuously repeat all at the speed of a computer.

ML is becoming very popular in software products and applications, and in all sectors from accounting to hot dog recognition apps. When any of these ML techniques are added to exciting projects, it leads to some difficulties.

ML often expands the purpose of applications that are in existence including web shop recommendations, classification of utterances in a chatbot, etc. It will be a part of the massive cycle with new additional features, fixing bugs or other reasons for frequent modifications in overall code.

Likewise, ML can exist in many ways in the next generation of Automation. DevOps with automation enables a rapid SDLC, but one that is too distributed, dynamic, opaque and ephemeral for comprehension of the human. Similar to automation, ML uniquely handles the volume, velocity, and variety of data that is generated using new delivery processes and using the next-generation of atomized, composable, and scaled out applications.

Machine learning

 

Some of the key examples of applying Machine Learning to DevOps include:

 
Tracking Application Delivery

The activity data from ‘DevOps tools’ such as Git, SonarQube, Jira, Ansible and many others provide delivery process visibility. Application of ML on these tools uncover the anomalies in that data- substantial code volumes, long build times, late code check-ins, slow release rates to identify many of the software development wastes, including excessive task switching, gold plating, inefficient resourcing, partial work, or process slowdowns.

Ensuring Application Quality

ML, by analyzing output from testing tools can intelligently review QA results, efficiently build a test pattern library based on discovery. This understanding of a ‘known good release’ helps to ensure comprehensive testing on every release, even for novel defects, increasing the quality of delivered applications.

Securing Application Delivery

Like fingerprints, User behavior patterns can be unique. Applying Machine Learning to Dev and Ops user behaviors helps in identifying anomalies, which represents dangerous activity. For instance, the access of anomalous patterns to repos, deployment activity, automation routines, test execution, system provision, and more can highlight users were exercising ‘familiar bad patterns’ in a rapid pace both intentionally or accidentally. These patterns include deploying unauthorized code, coding back doors, stealing intellectual property, etc.

Managing Production

Machine Learning comes into its own by analyzing an application in production, due to larger data volumes, transactions, etc. occurs in prod when compared to dev or test. The teams of DevOps use ML to analyze general patterns including resource utilization, user volumes, etc. and finally to detect abnormal patterns like memory leaks, DDOS conditions, and race conditions.

Managing Alert Storms

The practical and best-value application of ML is in managing the gigantic flood of alerts, which occurs in the systems of production. It can be more complicated such as “training systems overtime to identify ‘known well’ and inadequate warnings, thus enabling filtering to reduce alert storms and fatigue.

Troubleshooting and Triage Analytics

The other area where Ml technologies shine today is in triage analytics. It can automatically detect and can triage known issues and some unknown ones even. These tools can detect anomalies in general processing and can analyze the release logs to correlate with new deployments. Even other automation tools can use ML to raise a ticket, alert operations and assign them to the exact source.

Preventing Production Failures

In the prevention of failures, ML can go beyond the straight-line capacity planning. It can map utilizing patterns to predict. The required configuration for a desired level of performance, the percentage of clients can use a brand new feature, infrastructure necessities for a brand new promotion, an outage can impact the engagement of customers. ML sees opaque early indicators in applications and systems, allowing Ops to avoid problems faster with quick response times.

Analyzing Business Impact

In DevOps, to achieve success, understanding the impact of code release on business goals is critical. ML systems can detect good and bad patterns by analyzing the metrics of the user thus generate an early warning system to business teams and coders when a problem arises in applications.

 
Conclusion

Though ML has no short or easy button, and there is no alternate for intelligence, creativity, experience, and struggle. But today we see many of its applications and the sky is the limit as we continue to push the boundaries.

 
Bio: Prasanthi Korada is a Post Graduate in Computer Science from Andhra University, and was born and raised in Kakinada. She is currently working as a Content Contributor at Mindmajix.com. Her professional experience includes Content Writing at Vinutnaa IT
Services and Digital Marketing at Prominere Software Solutions. She can be contacted
at deviprasanthi7@gmail.com. Connect with her also on LinkedIn and Twitter.

Related:


Sign Up