2018 was a good year for enterprise IT with innovative technologies beginning to be adopted by the mainstream. We also began to have discussions about the next leap for IT operations, introducing the idea that artificial intelligence will soon play a major role. 2019 is where this use of AI and machine learning in operations will become more of a reality. Here’s what to expect.
Getty Images
1) Great Complexity Will Demand AI-Driven Solutions
The need for the support of AI comes from the fact that our IT environments are continuously becoming more complex. Monolithic applications living in a single data center are being replaced with microservices-based applications running in container systems hosted by multiple clouds. And as these services become more ephemeral it becomes impossible for humans to keep track of everything that’s going on. And yet understanding what applications are running where, and how that is impacting the business is critically important.
2) Multiple Tools Will Work Together to Deliver AIOps
Given the heterogeneity and size of modern IT environments, there will be no single tool that gathers the entirety of data available. Instead we will find data from multiple tools being combined into a unified data platform that will then be able to provide a set of powerful tools such as baselining and anomaly detection over massive amounts of metrics and event data. Machine learning will be used to correlate this data together, making problem identification much faster and therefore decreasing time to resolution when problems do occur.
3) We’ll Move Beyond Root Cause Analysis to Remediation
By combining a unified data platform with domain models that are specific to IT environments, we will begin to not only automatically identify the root cause of any problems, but be able to make suggestions as to how problems can be resolved. Incident response tools will capture the steps taken by operations teams and machine learning will begin to recognize where new events are similar to previous events, elevating the steps taken previously so that they can be quickly re-executed to address the current concern. While human approval will likely be required for most tasks, there may be specific situations where tools go so far as executing remediation steps on their own, similar to today’s auto-scaling features.
4) AIOps Will Impact More of the SDLC
The data from AIOps tools is useful for more than just production monitoring, the intelligence can be applied throughout the software development lifecycle. Data around resource consumption can be used not just for capacity planning but also to highlight where architectural adjustments may soon become necessary. Errors found in production can be catalogued so that developers will have automated suggestions on where they may be heading down a problematic path. An understanding of how users are exercising an application in production will be used to automatically generate regression tests to ensure that code changes will not break things once deployed. We can expect constant innovation in this space as we improve our data collection and as we learn to apply machine learning to more IT use cases.
5) Business Metrics Will Remain Critically Important
What must not change with this automated revolution is the understanding that the technical metrics do not matter without understanding their impact on the business. That is why it will be critical that business metrics are also fed into the operations platforms, so that movement in business KPIs can be correlated to changes in the technical environment. With this data, IT will focus on addressing the most impactful areas up-front, in turn driving higher customer satisfaction (both internal and external) and improving both the top and bottom line.
Business metrics matter greatly to enterprises. In a recent AppDynamics survey of global IT leaders, 74% of respondents said they want to use monitoring and analytics tools proactively to detect emerging issues that impact business performance, optimize user experience, and drive business outcomes like revenue and conversion.