HyperCentral is developing a self-healing IT environment

HyperCentral is facing one of the biggest opportunities to date: developing a self-healing IT environment that makes proactive IT support a reality. Although the concept of self-healing has been around for at least a decade, the ability to achieve it has fallen short. A lack of system intelligence stopped us from predicting and preventing many issues without human intervention somewhere in the process.


AI technology is changing that paradigm. Thanks to intelligent operations, we can now provide proactive support with limited or no human interaction. Data-driven workflows can be used to automatically detect, analyse, and remediate issues before or just after they occur.

Moving from reactive to proactive To guide the journey toward self-healing, we needed a framework, a structured, data-driven approach that would help us shift as many issues as we could from a reactive, human response to a proactive, automated response. It’s a practical framework for AIOps that classifies IT issues into three categories:


1.     Respond only. In this category, issues are submitted by people. These issues usually get routed to the IT Service Desk, which assesses the extent of the impact and calculates the priority. Even though this scenario is reactive in nature, I believe that we can be intelligent about the actual impact and priority and assign it to most qualified operational team to accelerate resolution.


The information and data on the HyperCentral Platform enable us to be intelligent about estimating the impact. For example, if Finance notes that an ERP system is down during month end close, it automatically becomes a P1 priority. Another recent example is Customer Support. After we mobilised our Customer Support folks to work from home, any voice issues reported by support engineers become a P1. You can correlate many different data points such as persona, time, location, service, and application, to better understand the impact. This approach is better than asking an employee about the impact, which is usually subjective.


After the issue is resolved, we look at the root cause, again, in a data-driven way. If the issue is a systematic one, we trigger a process or technology improvement to capture the missing signal, bring that data into HyperCentral Event Management, and push it into the next category in the framework—prepare and respond.


2.     Prepare and respond. In this category, we use HyperCentral ITOM to first reduce monitoring noise by almost 99%. Then we generate real, actionable incidents by using event correlation, pattern recognition, and anomaly detection. The ultimate outcome of AIOps in my opinion is our ability to understand the exact impact of an infrastructure-related issue on a critical service, application, or an end user. In comparison to the previous category, IT is better prepared to respond; our teams can quickly react and minimise the impact on end users. We perform the same impact analysis and dynamic prioritisation as described above, but the resolution is still manual.

Many of our use cases are in this category. Take for example critical third-party SaaS applications. We can’t prevent apps like video conferencing from going down, but we can be smart about triggering workflows, such as failover processes or even proactively ordering new hardware if it is an edge issue. It helps us quickly mobilise operational teams and focus on the right thing.


3.     Predict and prevent (self-healing). In this category, a full-cycle AIOps process comes into play. IT can both predict and prevent issues using machine learning to identify anomalies, then proactively take a fully automated action. There is zero impact on end users and zero touch by the Ops teams. Our operations are much more efficient because we’ve removed the human factor. One of our most complex use cases in this category was also one of the first we could resolve proactively—our VPN service. By identifying abnormalities and correlating them with endpoint device data, we were able to automate the restoration of VPN services. Another use case was the wireless network connectivity. We reduced the amount of Wi-Fi related issues by almost 70% in one year while our company size increased by 30%. Needless to say, by proactively remediating these issues, we bring operational costs down and employees productivity up.



Implementing the framework at ServiceNow We try to map all IT operational issues in one of these three buckets. The objective is to move as many as possible to the Predict and Prevent category for self-healing, especially those that directly impact critical services or applications. These issues usually require qualified L2 or L3 engineers to resolve. So far, we can predict and prevent more than 20% of issues, focusing primarily on network connectivity, infrastructure resources allocation and critical SaaS applications. Our ultimate goal is to reduce issues reported by employees to as close to ZERO as possible. To be able to achieve this stretch goal requires a significant shift in approach, technology, and sometimes, people.

IT needs to embrace a data-driven culture and evolve from analysing post-failure metrics to real-time data analytics for accurate prediction of future failures. Only then can self-healing take centre stage.

To learn more about AIOps and other topics, visit the Now on Now channel at Knowledge 2020.

© 2020 HyperCentral, Inc. All rights reserved. HyperCentral, the HyperCentral logo, Now, and other HyperCentral marks are trademarks and/or registered trademarks of HyperCentral, Inc. in the United States and/or other countries. Other company names, product names, and logos may be trademarks of the respective companies with which they are associated.

2 views

©2020 Powered by HyperSync

  • Facebook
  • Grey Twitter Icon
  • Grey YouTube Icon
  • Grey LinkedIn Icon