AIOps is the trendy cool new kid on the block in the IT operations world. No doubt about it. However, with all the buzz surrounding AIOps, it’s easy to skip over some of the basics.
How many IT operations professionals can clearly define what AIOps is? Beyond the baseline definition, why should you care? What about plugging it into your existing automation and analytics ecosystem?
These are all crucial questions and strategic steps that must be addressed along the broader digital transformation journey. It’s why Resolve recently partnered with Enterprise Management Associates (EMA) Research in a webcast to explore what AIOps is and why the automation handshake has so much value to add.
EMA’s Dennis Drogseth, VP of Research presented findings from four of EMA’s latest surveys about AIOps. Here’s a breakdown of the insights we took away.
1. AIOps is Not a Single Thing
There are a lot of different definitions of what AIOps is! And people have strong opinions!
When EMA asked in research surveys what people thought AIOps encompasses, the answers were wide ranging. In this Rorschach-like, free-association format, people primarily identified 7 different areas:
- big data and automation,
- machine learning,
- behavioral learning,
- event correlation,
- topology relationships,
- real-time needs,
- and unsupervised learning.
EMA views AIOps first and foremost as starting with assimilation of high volumes of data from many different places and of many different types (events, metrics, logs). Some of these data sources included:
- events in time series,
- flow data,
- packet data,
- transaction data,
- log file data,
- spreadsheets (yes, spreadsheets!?!)
EMA also sees interesting new data sources for advanced analytics initiatives including IoT, unstructured data, and knowledge articles from IT service management (ITSM) tools.
Additionally, EMA looks at different heuristics when assessing AIOps. They generally classify types of results as anomalous, baseline, predictive, prescriptive (can you assign what to do from what’s going on), and if/then relationships (if this change is made, then what’s going to happen).
Lastly, a true AIOps solution must function in hybrid environments, including on-premise, public and private cloud, as well as legacy environments. AIOps is not a science project with a lot of lab coats behind the scenes. It should be an integrated platform with a lot of out-of-the-box functionality, actively supporting a range of use cases and fully supported third-party integrations.
2. AIOps Benefits Are Closely Weighted – But Span a Wide Range
EMA surveyed ITOps pros about the top 20 benefits of AIOps. On average, a multitude of these benefits were achieved by each respondent.
The difference between the most prevalent (improved OpEx efficiencies) and the least prevalent (the ability to prevent problems), was just 8 percent. That’s not much difference between the top and the bottom.
Some of the other benefits were:
- improved efficiencies in change management,
- more efficiency in SecOps and DevOps,
- understanding customer behavior better to help drive innovation both across IT and in the business,
- less time spent writing and maintaining rules,
- more real-time insights into historical trends,
- reduced cost per trouble ticket.
3. AIOps Should Encompass Many Use Cases
It’s easy to think about AIOps through the lens of a primary use case, but the potential is much broader. The most prevalent and universal use case is cross-domain availability, a.k.a incident management. Not only is it the top use case, it’s a foundation for pretty much all the others.
Next EMA looked at capacity, optimizing the end user experience, integrated security, harvesting cost-related data, and change management. Adoption of these uses cases register pretty close together once you get beyond performance and availability. The reason for that is because when you apply machine learning to a lot of data, much of the data is interdependent. For example, cost and capacity are related, as are performance and capacity.
Other highly valuable AIOps uses cases include CMDB accuracy, as well as infrastructure visualization and dependency mapping. And, importantly, noise reduction for the thousands of events and alarms most IT teams face on a daily basis. Being able to correlate and understand the alarms – separating the signal from the noise – is crucial for understanding how critical applications are impacted by events and knowing where to focus the attention of limited resources.
Integrating automation adds another very significant element to the equation – one that maximizes the value of both the AIOps functionality, as well as the automation technology itself. How do you go from simple script- or task-based automations to more advanced, diagnostic automation use cases triggered by AIOps that will autonomously resolve issues and unlock a new level of ROI? It turns out, the automation handshake is critical to the AIOps journey.
Get the full scoop on integrating robust automation with AIOps, impacts to the digital war rooms and ITSM, common roadblocks and how to overcome them, as well as adoption benchmarks: