Check whether your instance fleet can carry peak internet egress through an internet gateway, including the single-flow limit.
An internet gateway is horizontally scaled and imposes no bandwidth limit of its own, so the real ceiling is the sum of the network allowances of the instances behind it. A second, easily missed ceiling applies per flow: traffic to the internet is capped at 5 Gbps for any single flow regardless of how fast the instance is, rising to 10 Gbps only inside a cluster placement group. Teams routinely blame the gateway when a single large transfer stalls at 5 Gbps, when the fix is parallel connections or a larger instance rather than a network redesign.
Internet Gateway Capacity
aggregate = instances × per-instance bandwidth; headroom = (aggregate − required) ÷ required; instances needed = ceil(required ÷ per-instance bandwidth).
aggregate = instances × per-instance bandwidth; headroom = (aggregate − required) ÷ required; instances needed = ceil(required ÷ per-instance bandwidth). An internet gateway is horizontally scaled and imposes no bandwidth limit of its own, so the real ceiling is the sum of the network allowances of the instances behind it. A second, easily missed ceiling applies per flow: traffic to the internet is capped at 5 Gbps for any single flow regardless of how fast the instance is, rising to 10 Gbps only inside a cluster placement group.
Teams routinely blame the gateway when a single large transfer stalls at 5 Gbps, when the fix is parallel connections or a larger instance rather than a network redesign.
This calculator takes 5 inputs: Peak internet egress required, Instances carrying internet traffic, Network bandwidth per instance, Largest single TCP flow, Single-flow limit that applies. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
No. AWS documents the internet gateway as horizontally scaled, redundant and highly available with no availability risk or bandwidth constraint. Throughput limits come from the instances, their network allowances and burst credits, and the per-flow cap.
Each flow — one 5-tuple of source IP, source port, destination IP, destination port and protocol — is limited to 5 Gbps to the internet. Multiplying the connections, as multipart S3 transfers do, is what unlocks the rest of the instance allowance.