Size a LAG: aggregate bandwidth, capacity after a member fails and the single-flow ceiling hashing imposes.
A LAG adds bandwidth but not per-flow speed: the load-balancing hash sends every packet of a flow down the same member to preserve ordering, so a single TCP session can never exceed one link. Sizing therefore needs both the aggregate figure and the post-failure figure. Bundles are routinely run above 50 percent, which looks fine until one member drops and the remainder tail-drops — the post-failure utilisation is the number that decides whether a fibre cut is an incident.
Port Channel Capacity
Aggregate = active members × member speed; after a failure it is (members − failures) × member speed, while any single flow is still capped at one member's speed.
Aggregate = active members × member speed; after a failure it is (members − failures) × member speed, while any single flow is still capped at one member's speed. A LAG adds bandwidth but not per-flow speed: the load-balancing hash sends every packet of a flow down the same member to preserve ordering, so a single TCP session can never exceed one link. Sizing therefore needs both the aggregate figure and the post-failure figure.
Bundles are routinely run above 50 percent, which looks fine until one member drops and the remainder tail-drops — the post-failure utilisation is the number that decides whether a fibre cut is an incident.
This calculator takes 5 inputs: Active member links, Speed per member, Maximum members the platform bundles, Members to survive losing, Peak offered load. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
Because hashing is per-flow. A single flow is capped at 10 Gbps; to fill the bundle you need at least as many flows as members, and ideally many more so the hash distributes evenly.