skip to main content

Ovum view


At its recent Perform 2018 user event, digital performance management vendor Dynatrace laid the foundation for its next generation of application performance management (APM) that it calls self-driving IT and NoOps. At its core is the use of artificial intelligence (AI) for automating the processing of vast amounts of metric data. Dynatrace has built its own AI engine, which it uses across all its technology, including auto-baselining, problem detection, root cause identification, and remediation. It also uses AWS conversational AI technology with its Davis chatbot. As IT environments become more complex to manage and troubleshoot, Ovum believes that using advanced automation such as AI will become essential, and Dynatrace is at the forefront of this approach.

Managing IT environment complexity in real time with AI

Microservices and containerization introduce dynamic and complex interactions to the IT environment, making traditional APM unable to cope with the deluge of data that is also transient, necessitating the use of, for example, a time series database.

Dynatrace’s new platform is designed to react in real time to incidents/events, and it uses this data to drive the AI system to identify these events and their causes. An important distinction that Dynatrace makes is that its solution makes better use of context to determine true causation, while older technology can only identify correlations that have higher false positives (false alerts due to spurious correlations).

Dynatrace’s OneAgent technology detects dependencies in production environments in real time, and it has married this with its AI engine. Transaction tracing is used to understand dependencies on a service and application level. This is combined with end-user monitoring on a transaction basis, as well as TCP/IP connection monitoring to understand infrastructure dependencies, and log file analysis that is linked to the process that generated the log file that therefore adds another depth of dependency know-how. These dependencies are chained and create a map of "reality" that the AI engine is then able to mine to discover true causality.

Dynatrace is using AI to fill in the missing link of causality, while using deterministic methods to build the basic model of dependencies. This is unlike other approaches that train the AI to learn the dependencies and that can therefore introduce spurious correlations. This blend of technologies is proving ideal for automating APM. Ovum is impressed with the ambition of Dynatrace in pushing the boundary of APM and making optimal use of AI.

For Dynatrace NoOps means greater automation

The idea of NoOps stems from not needing a person to sit in front of the dashboard to monitor the IT environment, and instead a machine does that chore and alerts operators when something unusual happens. In Dynatrace’s own internal operations center, there are no runbooks used, with everything that used to be in runbooks now automated in the APM platform.

Where Dynatrace is leading, other organizations are likely to follow, and the traditional job roles performed by IT staff could see another shift as DevOps, automation, and AI cut further into traditional operations work, displacing operations staff to perform higher-value work.


Further Reading

Trends to Watch 2018: Machine Intelligence, IT0014-003350 (Nov 2017).

Market Radar: Cloud-native Application Performance Management, IT0014-003329 (Oct 2017)


Michael Azoff, Principal Analyst, IT Infrastructure Solutions

[email protected]