Right-Sizing a New Campus Core
A regional hospital is consolidating three aging wiring closets into a new two-building campus. Each building will have 14 access switches with 48 GbE copper ports and 4 SFP+ uplinks. Peak measured access-to-core demand is 22 Gbps per building today, expected to triple within four years. The core must also carry 8 Gbps of inter-building traffic and terminate two 10 Gbps WAN links. The facilities team can only provide single-mode fiber between buildings and has limited rack space in each IDF. The CIO wants a redundant design with no single point of failure and asks whether a collapsed core or a dedicated core layer is justified.
1. Recommend a core topology (collapsed vs. dedicated) and justify it with capacity and redundancy reasoning.
Read the worked answer
Recommendation: dedicated two-tier core.
- Capacity: projected access demand per building is 22 Gbps × 3 ≈ 66 Gbps. A collapsed core would force those flows through the same pair of switches that also handle 8 Gbps inter-building and 20 Gbps of WAN termination. Dedicated core switches sized with at least 2× headroom (e.g., 100G-capable chassis or stacked high-density fixed switches) keep aggregation and WAN roles separate.
- Redundancy: two core switches with redundant single-mode fiber between buildings, MLAG/VSX or stacking, and dual-homed access uplinks remove single points of failure. A collapsed core in one IDF makes that IDF a chokepoint.
- Rack space: the constraint argues against adding a third tier, but a dedicated core is only two switches per building - the real space cost is in the IDFs, so verify uplink port counts before committing.
- WAN: terminating 2×10 Gbps on dedicated core ports avoids oversubscribing aggregation links.
A collapsed core would be defensible only if projected demand stayed under roughly 40 Gbps and WAN termination lived elsewhere.
2. The access switches only have 4×10G SFP+ uplinks. Explain how you would cable them to the core and what oversubscription you accept.
Read the worked answer
Use two uplinks per access switch, split across two core switches (or an MLAG pair), giving 20 Gbps per switch and 2:1 oversubscription against 48 GbE downlinks (48 Gbps). With 14 switches per building, that is 28 uplinks per building - plan core port counts accordingly (e.g., 32-port 100G line cards broken out, or 40G/100G uplinks on the access side if budget allows).
Accepted oversubscription: 2:1 at the access edge is standard for hospital endpoints; do not exceed 4:1 on links carrying imaging or VoIP. Document the assumption that peak is 22 Gbps today versus 280 Gbps of theoretical line rate - the design is deliberately oversubscribed but must be re-evaluated when the four-year projection is revisited.