AIOps — Greater Than the Sum of its Parts
It’s more than performance monitoring
If you’re reading this, chances are you’ve not only heard about AIOps but have also been inundated by companies claiming that their AIOps solution is your fast lane to digital transformation. Just a few years removed from Gartner coining the term, 2019 became The Year of AIOps Hype.
A healthy amount of this hype is warranted. However, one research firm said the global AIOps market is expected to grow from $2.55 billion in 2018 to $11 billion by 2023, at a compound annual growth rate of 34%. But some industry watchers are urging caution, as there isn’t a clear consensus on what AIOps is, let alone how it can benefit enterprises today. Meanwhile, an overwhelming number of companies are suddenly claiming that they are now an AIOps vendor — a reflection of the pressure companies are feeling to align themselves with the term.
The term “AIOps” stands for “artificial intelligence (AI) for IT operations,” and its premise is to give enterprise management teams a view into how IT is supporting their businesses. AIOps is described by Gartner as having two main components at its heart — big data and machine learning — which are supported by monitoring, service desk, and automation capabilities to give continuous insight into hybrid IT infrastructure. This insight can be incredibly valuable to enterprises, as it can help prevent business-impacting application outages and slowdowns — outcomes any organization wants to avoid at all costs.
The Evolution of AIOps
No AIOps company started that way — almost all of them evolved from their roots in performance monitoring, either for the network (NPM), infrastructure (IPM), applications (APM), or the service desk. Meanwhile, AIOps as a concept evolved from IT operations analytics (ITOA), which was an approach to IT operations data that allowed enterprises to understand and make decisions about their IT environments. But, while ITOA leveraged big data, it was missing the key elements that make AIOps so promising: AI and machine learning (ML).
As AIOps technology evolved, organizations continued to invest in APM, IPM, and NPM solutions to keep their IT teams informed of various useful metrics, such as uptime, software errors, transaction speeds, traffic statistics, processing, and more. Fast forward to today, and while many vendors still emphasize their roots in APM, IPM and NPM — since those solutions are still being widely used — AIOps is increasingly becoming the catch-all concept vendors are using to sell their solutions.
How Are Enterprises Using AIOps Today?
Today, most large enterprises use dozens of management and monitoring tools, usually as a combination of APM, NPM, and countless silo-specific IPM tools. But monitoring in particular is an issue in this scenario because the tools don’t relate to each other — meaning they essentially speak different languages. This means these tools can’t gain crucial context about what’s happening in other IT domains, making it very hard to determine the root causes of problems. These tools also have no inherent understanding of the applications running on the underlying hybrid IT infrastructure, nor the relative business value or importance of the applications.
Some organizations have implemented dedicated AIOps tools that simply collect and analyze alerts or analyze logs from all these other products, but these tools can’t be used for real-time performance monitoring or for proactive problem prevention. As a result, enterprises may think they’re making sense of the onslaught of data and alerts generated by the myriad of legacy tools by applying AIOps, but they’re really just executing basic event correlation and deduplication (or noise reduction) with some rudimentary pattern matching. This may help IT managers resolve problems after the fact, but these tools are very poor at preventing problems due to a lack of a real-time architecture for cross-domain correlation and analytics.
In terms of providing enterprises with the insight into the applications and underlying infrastructure they so desperately need, these narrow tools fall woefully short.
What Does True AIOps Look Like?
In order for enterprises to truly benefit from AIOps, a new, real-time, cross-silo, and application-centric approach is needed. Rather than looking at the user, application, and infrastructure independently, an AIOps platform should follow the journey of the application from the end user into the increasingly hybrid IT infrastructure and back out again. And rather than the journey being monitored in sections by the tool that’s most relevant at that point in the journey (APM, IPM, etc.), the AIOps platform should provide an end-to-end, app-centric view.
A comprehensive AIOps solution should also apply real-time, AI-based analytics featuring machine learning, statistical analysis, heuristics, and expert systems across the entire infrastructure. The solution should be designed for mission critical apps that need to be managed in real time with foresight through capabilities, such as workload infrastructure balancing and capacity forecasting. The modern AIOps platform will also suggest optimizations and ways to prevent problems. This enables enterprises to proactively identify and resolve issues that impact their business-critical applications, which, in turn, removes the need for costly reactive firefighting.
AIOps: Much More Than Glorified Monitoring
There is no single APM, IPM, NPM, service desk, or ITOA solution that can deliver the meaningful business impact outlined above, and that’s why AIOps is so much greater than the sum of its parts. By deploying a true AIOps strategy that leverages an integrated performance management ecosystem — not revamped, legacy tools — enterprises will be able to quickly understand whether the applications that are running their business are performing as they should. That’s the ultimate SLA, and true AIOps can ensure that enterprises meet it.