Real-Time Analytics Platform · View 03 of 21 · Structure

Layered Architecture

What depends on what. Dependencies point downward only; platform services are consumed by every layer above them.

Editable source SVG draw.io All views

Dependency rule

A layer may call the layer directly below it and never the one above
Consumption never reaches past serving into the analytics store; no direct cluster access for BI users
Metric semantics live in versioned KQL functions, not in dashboard definitions

Why layered at all

Lets the hot store be replaced without renegotiating consumption contracts
Makes the blast radius of a change explicit before it is made
Gives each layer a distinct owning team and a distinct on-call rota

Known tension

Latency-critical consumers read gold event topics directly, bypassing the serving layer
That shortcut is explicit, contract-governed and rate-limited, not accidental coupling
Any new bypass requires an architecture decision record, not a configuration change

◀ High-Level Architecture All views Container Architecture ▶