Five Steps to Accelerate Your AIOps Journey

Five Steps to Accelerate Your AIOps Journey

For today’s IT operations teams, understanding and solving issues can feel like a nonstop game of whack-a-mole. Managing a diverse, edge-powered infrastructure isn’t easy for increasingly complex, distributed enterprises. Most IT operations teams have organically accumulated multiple legacy tools as their organizations have matured, each with its own distinct interface and training requirements. This tools glut can drag down IT operations and limit their ability to gain visibility. Manual processes make the problem worse, because they make correlating data across massive amounts of data a slow, manual, and error-prone process. 

Even a relatively minor issue can send a help desk engineer deep into a frustrating journey. For example, if a user complains about an artifacting Zoom session, it’s up to the support engineer to first determine if they’re the right person to tackle the issue. If they have the right tools and visibility needed to make a diagnosis, they will still need to investigate the cause. Is the problem with employee’s edge device, their application, wireless access point, switch, or the WAN itself? Failure to accurately chase down the root cause leads to finger pointing between teams—and wastes time and money. 

Network environments are growing fast, and it’s clear that traditional approaches to management can’t keep up. To break through the old limitations, enterprises need to automate diagnostic flows to proactively recommend solutions to emerging problems. 

A proactive approach to infrastructure health

AIOps brings together automation and artificial intelligence machine learning (AI/ML), enabling IT to focus on implementing preventative mechanisms, instead of reacting to issues. 

AIOps is fundamentally about building on visibility, so an effective solution will extend from the WAN to the edge, and all points in between. An SD-WAN solution can deliver great visibility into network activity between different branches and cloud applications. To get the full picture, LAN visibility is key for today’s edge-driven environments—especially for challenging IoT use cases like retail sites using wireless point-of-sale devices. An effective solution should not only deliver visibility into what’s happening on the LAN, but apply AI/ML to help organizations surface the most relevant data and cross-correlate it. 

AIOps excels at identifying the root cause of issues and providing recommendations on how to resolve them. However, a successful implementation requires some planning and alignment with your specific environment and processes. Here are five steps to help you move forward on your journey.

1) Catalog your data sources

You can’t enhance observability and drive better decisions until you understand the full scope of your organization’s data. Take time to determine which data elements you will need to collect—and what those elements mean in terms of impact to your environment. Consider the data format, who can access it, and who retains ownership of the access.  

2) Collect required data

After cataloging your various data sources, you can move into the collection phase. Determine how to access the data you need, and which protocols are required to do it. Collecting data is not limited only to statistics and logs; you will also want to consider your organization’s topologies to gain context around where your data is most relevant. You should also evaluate systems that interact with your data, such as ticketing systems. 

3) Correlate data

Once you’ve collected your data, the next step is to normalize and correlate it.  AI/ML engines can really shine in this process. You will want to examine data across multiple functional domains, and even across multiple vendors. The goal is to start pulling some of these pieces together so that you can establish a baseline and extract what’s most relevant. 

4) Root Cause Analysis

After establishing a baseline for what is considered normal behavior in your environment, you can take steps to determine the root cause of a manifesting issue, using anomaly detection. 

5) Correct issues

After building an understanding the root causes of specific issues, you can take action to correct them. AIOps can not only deliver mitigation recommendations but automate specific actions—with a clear set of rollback criteria so the risk is minimal. 

As you embark on your AIOps journey, start with the answers you want the solution to produce, and keep the scope contained. It’s an interactive process of working your way backward, identify the signal you need, and the data support it—and know how to access the data. Although AIOps can’t happen overnight, the result will be a more agile, proactive approach to managing your environment.

 

Author's Bio

Tim Van Herck

Director Technical Product Management
VMware