Dynatrace Platform Fundamentals
The three tiers
Every Dynatrace deployment is described by three tiers: OneAgent collects data on the host; ActiveGate routes traffic, executes extensions, runs private synthetic and, in Managed clusters, brokers the connection to Dynatrace; the cluster (SaaS, Managed, or a self-hosted cluster) stores, models, and analyses everything. OneAgent is not "installed per application" - it is installed once per host and injects code modules into supported runtimes automatically.
The entity model
Collected signals become entities in Smartscape - hosts, process groups, processes, services, applications, Kubernetes objects, cloud services - joined by typed relationships. Almost every exam question about "why did Dynatrace show this?" resolves to entity topology. Grail is the data lakehouse behind the platform (logs, events, spans, metrics, security events), queried with DQL and governed by buckets with retention rules. Davis is the causal AI: it walks the topology to merge correlated events into a single problem with a root cause, rather than firing per-metric alerts.
Segmentation primitives
Tags (key:value, automatic or manual) are for filtering and reporting across dashboards, metrics, and DQL. Management zones are for scoping access, alerting profiles, and dashboards to a team. IAM policies with boundaries grant permissions. Confusing tags with management zones is the single most common trap.
Worked rule of thumb
Ask: does the requirement mention access restriction or team isolation? → management zone. Does it mention filtering a dashboard or metric query by environment or owner? → tag. Does it mention automating a remediation step? → workflow / AutomationEngine, triggered by a problem event.
Other traps: ActiveGate does not collect host CPU/memory metrics (that is OneAgent); Davis does not train on your data; a problem is not the same thing as an event.