MTU / VLAN (not modelled)
MTU is how big a single chunk of data is allowed to be — "jumbo frames" are simply larger chunks, so the network spends less time on packaging and more on payload. A VLAN is a way to slice one physical switch into several logical networks that can't see each other, so two tenants sharing a box stay walled off. Both matter enormously in a real datacenter.
Neither one moves the dial in this game. You won't find an MTU slider to tune or a VLAN to carve up, and you don't need to — nothing in the simulation rewards or punishes you for them. If a guide tells you to "enable jumbo frames for performance," that advice is about a real network, not this one. Spend your attention on the things that ARE modelled: fabric capacity, oversubscription, and latency.
Detailed explanation
Real-world throughput tuning leans hard on a 9000-byte jumbo MTU (fewer per-packet interrupts, higher goodput) and on VLAN isolation for multi-tenant segmentation. The engine tracks neither: there is no mtu field on Port and no 802.1Q tag on Cable. Reachability is a connectivity sweep, not an L2 broadcast-domain model, so VLAN boundaries simply don't exist in the data model.
This is the one operations topic with no grounded fact behind it — there's no live state to probe because there's nothing to measure. Treat MTU/VLAN as flavour, not strategy. What IS modelled and worth your time: fabric_capacity_gbps (the ASIC ceiling), the uplink:downlink oversubscription ratio, and the per-tier p99 latency budget. Tune those.