Real-Time Analytics Platform · View 02 of 21 · Context and scope

High-Level Architecture

The end-to-end shape in one picture: managed ingestion into an Event Hubs durable log, Databricks stream processing, and Azure Data Explorer as the hot analytics store behind a governed query surface.

Editable source SVG draw.io All views

Design stance

One durable log for every producer type — no parallel integration estate to operate
One stream processing engine, not two; simplicity of operation beats per-case optimality
Hot store is purpose-built for time-series analytics rather than a general-purpose warehouse

Latency budget · 5 s p95

Producer to 202 Accepted: 200 ms · collector publish to Event Hubs commit: 80 ms
Micro-batch trigger 1 s, enrichment and window 1.2 s, ADX ingestion visibility 1.5 s
Total 4.4 s at p95, leaving roughly 600 ms of headroom against the stated target

Rejected alternatives

Azure Stream Analytics as the only engine — session windows and complex joins do not fit
Fabric Real-Time Intelligence end to end — attractive, but couples ingest and BI licensing
Direct ADX ingestion with update policies only — no place for reference-data joins at scale

◀ System Context All views Layered Architecture ▶