Cross-AZ Traffic Costs Just Got a Default Fix
Amazon ECS Service Connect now supports zone-aware routing, announced July 23, 2026. The feature is on by default for every new and existing service. It costs nothing extra.
This matters because multi-AZ ECS deployments have always carried a quiet tax. Every service call that crossed an availability zone added latency and incurred additional data transfer charges. Teams either accepted that cost or constrained their topology, giving up resilience to save money. Zone-aware routing removes that trade-off directly.
What the Routing Actually Does
Each ECS task runs an Envoy sidecar proxy that handles service discovery, load balancing, and zone-aware routing. The proxy now preferentially routes requests to endpoints in the same AZ as the originating task.
It does not simply pin traffic to one zone. The algorithm compares endpoint distribution percentages between source and destination clusters in each AZ. Surplus capacity in a destination zone absorbs traffic from overloaded zones. When local endpoints become unhealthy or fall below capacity thresholds, traffic automatically redistributes across healthy AZs. No single zone gets overwhelmed.
As endpoints scale up or down, routing decisions update in real time without requiring redeployment. The system is self-correcting.
The Numbers in the Source
AWS claims more than 80% of traffic stays within the same AZ when endpoints are balanced across AZs. The blog post reports approximately a 24% reduction in median network latency in multi-AZ deployments when endpoints are balanced. Intra-AZ latency runs as low as 300 to 400 microseconds, compared to 1.5 milliseconds or more for cross-AZ calls.
These figures assume balanced endpoint distribution. Unbalanced clusters or uneven scaling may produce different results. The source does not quantify savings for skewed deployments.
Who Should Act and How
Cloud platform owners and application teams running ECS services across availability zones should move now. The action is low-risk: no application code changes and no infrastructure changes are required.
The one catch: existing services require a one-time redeployment to enable the new routing behavior. New services pick it up automatically. A redeployment triggers the Envoy proxy update in each task. Until a service is redeployed, it continues routing the old way.
The feature is available in all AWS commercial and AWS GovCloud (US) Regions where ECS Service Connect is supported. Teams running in GovCloud face no regional delay.
Validating the Change
Deploy first, then measure. Amazon VPC Flow Logs with AZ metadata can monitor cross-AZ traffic patterns and validate routing effectiveness. A drop in cross-zone byte counts after redeployment confirms the proxy is routing locally.
Watch for AZ imbalance during scale events. The algorithm redistributes load automatically, but a zone losing most of its tasks will push traffic across AZ boundaries by design. That is the failover working correctly, not a misconfiguration.
The decision now on the table: schedule redeployments for existing ECS services, set a Flow Logs baseline before redeployment, and compare cross-AZ traffic volume afterward. Teams with data-intensive workloads stand to see the largest cost reduction. How large depends on current traffic volume and AZ distribution, neither of which AWS quantifies in the announcement.