Real-Time Analytics Platform · View 16 of 21 · Operations

Deployment and Infrastructure

What runs where, which failure domains it spans, and what is actually standing by in the secondary region.

Editable source SVG draw.io All views

Availability target

99.9% monthly, allowing roughly 43 minutes of downtime — met by zone redundancy, not by DR
Event Hubs, ADX, ADLS and Redis are all zone-redundant within the primary region
A single zone loss is a capacity event, not an outage; the pipeline continues degraded

Disaster recovery

RTO 60 minutes, RPO 5 minutes — warm standby, capacity provisioned on declare
Event Hubs geo-DR replicates metadata only; in-flight events within RPO can be lost
Bronze is geo-replicated, so a regional failure costs a replay window and not the data

Honest gap

A regional failover has not been rehearsed end to end; the RTO figure is a design estimate
Quarterly game-day exercise is a prerequisite for production sign-off
Cross-region egress cost during failover is not yet modelled

◀ Replay, Backfill and Correction All views CI/CD and Environment Promotion ▶