AIOps relationship to MLOps, DevOps, and Site Reliability
Domain 3 - AIOps relationship to MLOps, DevOps, and Site Reliability
This domain tests whether you can place AIOps among adjacent disciplines without collapsing them into synonyms. Questions often ask who owns a practice, which feedback loop is described, or how AIOps supports SLOs.
Definitions that must stay distinct
- DevOps is a culture and set of practices that shorten the delivery feedback loop through collaboration, CI/CD, and automation.
- SRE applies software engineering to operations, using SLOs, error budgets, toil reduction, and blameless postmortems.
- MLOps manages the ML model lifecycle: data versioning, experiment tracking, model registry, deployment, monitoring for drift, and retraining.
- AIOps is the application of big data and ML to IT operations data to improve detection, correlation, and remediation.
AIOps is a consumer of MLOps practices when it builds its own models, and a provider to DevOps and SRE when it reduces noise and MTTR. MLOps may also consume AIOps telemetry to monitor production models.
How the feedback loops connect
DevOps produces change events and deployment telemetry. SRE defines reliability targets and decides when to slow releases. AIOps correlates those changes with incidents and surfaces SLO burn. MLOps governs the models AIOps uses, including drift detection and rollback.
Worked example
An SRE team sets a 99.9% availability SLO. After a Friday deployment, the error budget burns four times faster than normal. AIOps correlates the burn with a configuration change, a latency shift in one service, and a spike in 5xx logs, then opens an incident with the likely change attached. The team rolls back. MLOps later retrains the anomaly model with the incident labeled as a true positive. Reasoning: DevOps supplied the change record, SRE supplied the SLO threshold, AIOps performed correlation, and MLOps governed the model update. No single discipline owned the outcome.
Traps that cost marks
- Treating AIOps as a synonym for MLOps; MLOps is about models, AIOps is about operations outcomes.
- Claiming AIOps replaces SRE; it supports SLO decisions but does not set them.
- Assuming AIOps is a DevOps tool category; it spans monitoring, ITSM, and automation.
- Saying MLOps only applies to AIOps platforms; it applies to any production model.
- Missing that AIOps models themselves need MLOps governance.