Company is expanding into a second region for latency reasons. Data pipelines currently run entirely in one region. Designing the multi-region setup, active-active processing in both regions, or active-passive with failover?
Active-passive is simpler to reason about and usually sufficient for data pipelines specifically, as opposed to user-facing request serving. Most pipeline workloads tolerate the failover latency of promoting a passive region during an outage, and it avoids the conflict-resolution complexity of active-active writes to the same datasets.
Active-active earns its complexity when regional latency for pipeline inputs genuinely matters, say ingesting region-local IoT data that shouldn't cross regions before processing. Call out the specific requirement driving it rather than defaulting to it for general resilience.
Sign in to reply.
