How Co-locating App Pods With Stateful Backends Can Cut Cloud Costs

Cloud providers charge for cross-availability-zone (cross-AZ) data transfers even when traffic stays within the same region, creating significant ongoing costs for data-heavy platforms. When stateful workloads like databases, caches, or message queues run in a single AZ, spreading stateless application pods across multiple AZs offers no real availability benefit if the backend itself is not replicated. In such setups, all application pods still fail if the AZ hosting the sole stateful backend goes down, meaning the cross-AZ spread of app pods only adds transfer costs without improving resilience. Co-locating stateless pods in the same AZ as their stateful anchor removes this wasted expenditure without meaningfully reducing availability. Tools like Kyverno can automate this co-location by dynamically injecting AZ constraints at pod admission time, avoiding the error-prone process of manually managing node affinity rules across multiple workloads.