Real-Time Analytics Platform · View 04 of 21 · Structure

Container Architecture

The deployable units, the technology each runs on, and the protocol on every hop — what an engineering team actually builds and owns.

Editable source SVG draw.io All views

Component decisions

Event Collector on Container Apps rather than Functions: long-lived AMQP connections and KEDA scaling
Debezium on Container Apps rather than Data Factory: log-based CDC with sub-second latency
Event Hubs Dedicated rather than Premium: 3.2 billion events/day is past the Premium ceiling

Network posture

Public network access disabled on Event Hubs, ADLS, Key Vault, ADX and Redis
Every hop uses a private endpoint and a managed identity; no connection strings anywhere
Databricks is VNet-injected with secure cluster connectivity and no public IP

Deliberate omissions

Follower cluster topology and workload groups are drawn in the query and serving view
Retry, dead-letter and backpressure paths are drawn in the stream processing view
Availability zones and DR region are drawn in the deployment view

◀ Layered Architecture All views Integration and Interface Catalogue ▶