16. Topology visualization and analytical charts
17. Detailed results, bill of materials, and port allocation
Growth forecast
| Year | Endpoints | Demand | Required leaves | Required spines | Demand ratio | Status |
|---|
Bill of materials
| Category | Item | Quantity | Unit cost | Extended cost | Notes |
|---|
Port allocation
| Role | Devices | Ports per device | Used | Reserved | Free | Utilization |
|---|
Example data table
| Design | Leaves | Spines | Downlinks | Uplinks | Active downlink | Uplink | Ratio |
|---|---|---|---|---|---|---|---|
| Compact enterprise | 8 | 2 | 48 × 10G | 2 × 100G | 384 Gbps | 200 Gbps | 1.92:1 |
| General compute | 16 | 4 | 48 × 25G | 4 × 100G | 960 Gbps | 400 Gbps | 2.40:1 |
| Storage intensive | 24 | 8 | 32 × 100G | 8 × 400G | 2.56 Tbps | 3.20 Tbps | 0.80:1 |
| AI training pod | 32 | 16 | 32 × 400G | 16 × 400G | 10.24 Tbps | 6.40 Tbps | 1.60:1 |
Formula used
How to use this calculator
- Enter the current and future topology size.
- Provide leaf and spine platform specifications.
- Describe connected endpoints and active ports.
- Set traffic, utilization, and burst assumptions.
- Select failures and redundancy targets.
- Enter cabling, power, rack, and cost assumptions.
- Review warnings, charts, forecasts, and materials.
- Save scenarios and compare alternative designs.
Spine–leaf aggregation planning guide
Spine–leaf fabrics create predictable paths between racks and services. Every leaf normally reaches every spine. This structure supports equal-cost forwarding across parallel links.
Capacity planning begins with port counts and port speeds. Downlink bandwidth represents connected servers and appliances. Uplink bandwidth carries traffic leaving each leaf switch.
Oversubscription compares active downlink capacity against available uplink capacity. A higher ratio allows more endpoints per uplink. It also increases congestion risk during simultaneous demand.
Traffic assumptions matter as much as installed bandwidth. Many flows remain local to one leaf. Cross-leaf flows consume fabric capacity through a spine.
Storage, backup, and AI workloads can create synchronized bursts. Replication can multiply the original traffic demand. Elephant flows may also produce uneven ECMP utilization.
Failure analysis exposes designs that look healthy during normal operation. Losing one spine reduces path diversity and available capacity. Maintenance events can create similar temporary conditions.
Cabling calculations should include both endpoints of every link. Spare optics and cables reduce replacement delays. Routing allowance and slack prevent underestimated cable lengths.
Power calculations include switches, optics, and active NIC ports. PUE converts IT load into facility energy demand. Annual electricity cost becomes part of operating expenditure.
Growth forecasting identifies when ports or bandwidth become constrained. Endpoint growth and bandwidth growth rarely move equally. Future uplink speeds can change the preferred expansion path.
Advanced overlays add route, VNI, VRF, and session requirements. RoCE deployments also require careful congestion engineering. Vendor limits should always be verified before purchasing.
Frequently asked questions
What does this calculator size?
It sizes topology, ports, links, bandwidth, traffic, resilience, latency, optics, power, cost, growth, and advanced fabric scale.
How is oversubscription calculated?
Active server-facing bandwidth is divided by available leaf uplink bandwidth.
What is a non-blocking design?
A non-blocking design provides at least equal usable uplink and active downlink capacity.
Why model protocol efficiency?
Encapsulation, hashing, and path imbalance reduce useful throughput below physical line rate.
Why is failure-state analysis important?
Failures remove paths and concentrate traffic on surviving spine and uplink resources.
Does every leaf connect to every spine?
A standard two-tier fabric normally uses complete leaf-to-spine connectivity.
Can mixed leaf roles be represented?
Separate counts for server, border, and service leaves support role-aware planning.
Can the tool estimate optics and cables?
It calculates fabric links, server cables, optics, spares, fiber pairs, and patch panels.
Can it model GPU and RoCE networks?
High-speed links, rail-optimized mode, RoCE share, PFC, ECN, and lossless options are included.
How are costs calculated?
Capital costs and inflation-adjusted annual operating costs form the selected TCO period.
Where are saved scenarios stored?
Scenarios use browser local storage and remain on the current browser profile.
Can results be exported?
The page provides CSV, JSON, PDF, print, copy, and comparison outputs.
Are hardware presets authoritative?
No. Enter verified platform specifications from the selected manufacturer.
Should this replace a network design review?
No. Use it for planning, then validate protocols, optics, licensing, and operational requirements.