1 (866) 447-2526 Resources Events Blog

Machine Learning Analytics (Part 3)

Blog

Machine Learning Analytics (Part 3)


 

This article is part of a series covering how changes still present a risk for today's IT operations, despite advances in technology and processes, and how a change-centric analytics approach addresses this.

-  -  -  -

We’re talking about a lot of data. Just pulling various types of information from an operating system, like LINUX, can bring between 2 to 3000 configuration parameters. Going up the stack, like with data base application servers, there’s more and more information coming in, with thousands of granular changes and tens of thousands changes across the environment happening on a daily basis.

It’s essential to apply analytics to this data in order to make sense of it, and to make it actionable, bringing things that need to be dealt with to the attention of those that can make a difference.

Evolven’s Machine Learning Analytics Key Features

Evolven’s machine learning analytics features a number of key functions for making sense of all this data.

  • Noise reduction: Evolven automatically reduces noise and excludes all the irrelevant, noisy and low significance changes and differences.
  • Correlation: Evolven correlates data within all the data’s sources we bring and between multiple data sources.
  • Risk calculation: Evolven calculates risk scores for each change and difference dimension. A dimension, as we define this is kind of an algorithm or property of a change and difference.
  • Multi-dimensional risk analysis: Evolven combines all risk scores into the total risk, prioritizing changes and differences. This allows Evolven address questions:
    • Which change is the most probable root cause of an incident?
    • Which one is the most probable root cause of a future issue?
  • Automatic insights: Evolven translates the results of the analysis into automatic insights. An insight that is in natural language that explains the changes identified as a high risk. Evolven clusters related changes, explains their risks and points to the potential issues they can cause. An Evolven end-user can focus on the insights that tell this user where to invest their time and what to do to best address an issue.

Analytics Algorithms Calculate Risk

For the analytics algorithms that help Evolven calculate risk levels, probability of failure, there is a mix of machine learning statistics and heuristic algorithms. Heuristics looks at consistency, for example, like how consistently is a change deployed for all similar systems. For the case of applying a patch to five load balance servers, heuristics considers whether this patch was applied equally on all the servers. If it wasn’t, that’s an opportunity for an operational issue to occur.

Machine Learning Analytics Examples

An example for Machine Learning analytics would be to look at a time anomaly, like in a history of changes. In the case where infrastructure changes are always carried out between 11pm and 3am, then in the middle of the day a change in infrastructure of the same type is suddenly noticed. This would either be an emergency patch, authorized change, or an unauthorized change. In both cases the risk of such emergency activity is higher than changes carried out in regular maintenance.

Looking at the value of the change can provide insight. Usually this parameter would change from 1 to 10 or from 10 to 20. That’s the kind of pace that this change maintains. Suddenly when the parameter is shown to have changed from 1 to a value of 10,000, then it clearly appears like a fat finger mistake. This parameter was never changed this way before. This stands out as a suspicious change, making the risk from this change higher.

Evolven’s machine learning analytics goes through all the data selected by Evolven and also blended with data from other sources, when possible. Through this analysis, Evolven can point to the high risk changes and differences that can cause the trouble. Isn’t that the goal of analytics?

See Evolven in action!
Unlock the power of actual changes. Register now for a live demo.

 

About the Author
Sasha Gilenson
Sasha Gilenson enjoyed a long and successful career at Mercury Interactive (acquired by HP), having led the company's QA organization, participating in establishing Mercury's Software as a Service (SaaS), as well as leading a Business Unit in Europe and Asia.

Sasha played a key role in the development of Mercury's worldwide Business Technology Optimization (BTO) strategy and drove field operations of the Wireless Business Unit, all while taking on the duties as the Mercury's top "guru" in quality processes and IT practices domain. In this capacity, Sasha has advised numerous Fortune 500 companies on technology and process optimization, and in turn, acquired a comprehensive and rare knowledge of the market and industry practices.

Sasha holds an M.Sc. in Computer Science from Latvian University and MBA from London Business School.