AIOps: How To Get Started | 7wData

AIOps is about leveraging AI (Artificial Intelligence) and ML (Machine Learning) to automate various parts of an organization’s IT operations. “This provides modern ITOps teams a real-time understanding of any type of issues,” said Venugopala Chalamala, who is the founder and CEO of Atlas. “Traditional IT management solutions can’t keep up with the volume as well as provide real-time insight and predictive analysis.”

The need for AIOps has accelerated because of the growing complexities of IT systems, the explosive growth of data and the sudden increase in remote working arrangements. Gartner forecasts that the exclusive use of tools for this category will go from 5% of large enterprises in 2018 to 30% by 2023. 

So then if your organization is looking at AIOps, how do you get started and what are some of the strategies to consider? Well, to see, I have reached reached out to various tech experts to get their advice:

Clearly define the problem you want AIOps to solve. Is the goal to detect anomalies that are hard to find by a human? Or do you want a tool to enable your OPS team to identify root causes quickly when an issue occurs? Or do you want to deploy some automatic recovery mechanism through AI? AIOps can help in many different areas. This means you need to define a clear goal that will help you understand the potential ROI (Return On Investment).

Rosaria Silipo, Ph.D., who is the principal data scientist atKNIME:

You need a good understanding of what is necessary to monitor and store. The more AI models, the more complex the monitoring strategy. Then, you need to define the criteria of acceptable performances by a model or a group of models. Finally, a strategy is needed to retrigger training when performance drops below an acceptance threshold.

The value of an AIOps tool increases with the broad range of data that you can observe and analyze. It is also important that there is an open approach that can integrate with your existing IT tools and data sources. Once you have your tools, identify the right processes that support agility and collaboration across functions to integrate across Dev, Ops, and security. Finally, organizations have to think about the people–redeploy your most valuable resource to ensure the right tools and processes are in place and you can act on insights.

The key is to have a good incident management system. You also need to have a very good logging system in place. Also, there should be proactive and predictive management of incidents and outages. You don't want humans doing this.

When it comes to something as transformative as AIOps, start small. Choose a low-scale test case, learn, adapt, tweak, and grow from there. That way, if things go wrong, the consequences won’t be quite so disastrous.

Look for an AIOps platform that can perform automated procedures based on analytics drawn from your pools of data. Oftentimes, this data is already housed within your organization’s monitoring solutions. Then ask if the platform has dynamic thresholds, root cause analysis, forecasting and anomaly detection capabilities.

Eric Tyree, the head of AI and Research atBlue Prism:

AI is easy, Ops is hard: AIOps is all about automation, so make sure you are thinking about the whole automation toolbox. Mature automation programs should look to achieve a formula along the lines of 1/3 systems (AI, BPM, straight through processing), 1/3 human and 1/3 Digital Workers.