Enter for the full results page · Esc to close
One 800G port can serve one 800G device, two 400G, or eight 100G — and each choice needs a different module at both ends and a different cable in between. Pick the port, pick the ratio, and this gives you the bill of materials with the fibre count and the links to the parts.
What a breakout actually does. A high-speed port is not one signal — it is a group of electrical lanes. An 800G OSFP port is eight lanes of 100G. Breakout takes those lanes and points them at different destinations instead of one: eight lanes to one 800G device, four lanes each to two 400G devices, or one lane each to eight 100G devices. The cable re-groups the fibres; the switch OS re-groups the lanes.
Why the ratio has to divide the lane count. You cannot break a 100G QSFP28 into three, because it has four lanes. The valid ratios are the ones that divide the lane count evenly: 4-lane ports break into 1, 2 or 4; 8-lane ports into 1, 2, 4 or 8. Anything else strands lanes and most switch operating systems will simply refuse the configuration.
Configure the switch before you cable. Breakout is a port-level setting and on most platforms changing it needs a reload — sometimes of the whole switch. Decide the ratio at design time. Discovering after the cable is dressed that the port has to be reconfigured, and that the reconfiguration takes the switch down, is a bad afternoon.
Copper, active, or optical. Inside the rack use passive DAC: cheapest, lowest power, no optics to fail. Across a row up to about 30 m an AOC keeps the simplicity without the bulk. Beyond that you need parallel optics and structured fibre, which is where the MPO trunk and conversion harness come in and where the cabling planner takes over.
Usually yes, but often not within the same port group. Many ASICs configure breakout per group of four physical ports, so setting one port to 4×100G may force its three neighbours to match. Check the platform's port-group rules before assuming you can mix freely.
No. Breakout is defined by the IEEE lane structure and the MSA form factor, so a compliant module at either end interoperates. What matters is that the reach class and the wavelength plan match — a DR4 switch module needs DR-class host modules, not FR4.
In a 1:1 configuration, none go unused. In a breakout, the fibres are re-grouped rather than wasted — that is the conversion harness's job. Where fibres do get stranded is base-12 MPO trunk plant feeding 8-lane optics, which costs you a third of the trunk. The cabling planner quantifies that.
Turn a switch port count into the trunks, cassettes and harnesses you actually have to buy
Open tool →Pick a standard and get the wavelength plan, fibre type, connector, reach and planning loss budget
Open tool →Per-port wattage rolled up to switch, rack and cooling load — and whether the rack can take it
Open tool →Transceivers, MPO assemblies, WDM and patch cords — built and tested to the figures you just calculated.