AWS Cloud WAN: Where a Region's Egress Decision Is Made
AWS Cloud WAN is a managed core network that joins Regions, VPCs, and on-premises sites through attachments. This architecture view shows where the per-Region egress decision is made and how one link can be drained safely.
AWS Cloud WAN is a managed core network that joins AWS Regions, VPC attachments, and on-premises sites through a single policy-governed backbone. Each Region has a core network edge that chooses the egress path, and in a hybrid setup that choice follows the preserved on-premises AS_PATH rather than Direct Connect BGP communities. That distinction matters during maintenance: prepending the AS_PATH on one Direct Connect link steers only the affected Region to a peer path across the core, while every other Region keeps its closest exit. The diagram maps those parts and the one decision point engineers actually tune.
FAQ
Why does a Cloud WAN core edge ignore Direct Connect BGP communities?
In DeepL's hybrid setup, each core network edge makes its egress decision from the preserved on-premises AS_PATH, not from Direct Connect BGP communities. That is why prepending the AS_PATH on one link changes the decision for the affected Region only. Other Regions keep their local exit because their paths remain the shortest.
How do you drain one Direct Connect link without moving every Region's egress?
Apply an AS_PATH prepend on the link being maintained. Only the affected Region's core edge then prefers a peer path and crosses the core network to another Direct Connect location, while the remaining Regions keep their closest exit.
What do Cloud WAN routing policies add over per-link BGP tuning?
DeepL plans to adopt AWS Cloud WAN routing policies for deterministic failover and route summarization. That moves intent out of hand-tuned per-link BGP manipulation and into a declarative policy on the core network, which is easier to reason about during planned maintenance.
A layered architecture view of fine-tuning a 3D generation model: which flow DiT branches get LoRA adapters, which stay frozen, and how data, compute, and evaluation wrap the adapted core.
ARC Region Switch is a managed failover orchestrator that runs team-owned plans in dependency order. This diagram maps its trigger, orchestration, execution, and region layers and the ownership boundary.
Active learning is a loop that spends a fixed labeling budget on the samples most worth annotating. This diagram maps its stages — pool, score, label, train, evaluate — and the return path that closes the cycle.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.