Content Types
Select the Microsoft content types your organization will cache. This helps estimate per-device content volume.
Sites
Add each physical location that will have a Connected Cache node. Configure device counts, bandwidth, and preferences per site.
Sizing Results
Methodology
This tool combines official Microsoft guidance with reasonable assumptions to produce sizing recommendations. Below is a transparent breakdown of what comes from documentation and what is estimated.
Hardware Specifications MS Learn
The hardware recommendations (CPU, RAM, disk, NIC) come directly from the MCC prerequisites page. Microsoft provides three tiers of specifications. This tool maps them to site sizes as follows:
| Tier 1 → Branch Office (< 50 devices) | 4 cores, 8 GB RAM, 100 GB disk, 1 Gbps NIC |
| Tier 2 → Small-Medium (50–500 devices) | 8 cores, 16 GB RAM, 500 GB disk, 5 Gbps NIC |
| Tier 3 → Medium-Large (500–5,000 devices) | 16 cores, 32 GB RAM, 2x 200–500 GB disk, 10 Gbps NIC |
Note: The MS Learn hardware table does not explicitly label each column by site size. The mapping above is inferred by correlating the hardware table with the site category descriptions and bandwidth table on the overview page.
Bandwidth-to-Throughput MS Learn
The bandwidth-to-monthly-throughput values come directly from the overview page:
| 50 Mbps → 180 GB/mo | 100 Mbps → 360 GB/mo | 250 Mbps → 900 GB/mo |
| 500 Mbps → 1,800 GB/mo | 1 Gbps → 3,600 GB/mo | 3 Gbps → 10,800 GB/mo |
| 5 Gbps → 18,000 GB/mo | 9 Gbps → 32,400 GB/mo |
When your bandwidth falls between these values, we use linear interpolation. This is our calculation, not a documented method.
Multi-Node Sizing Assumption
Microsoft's largest documented site tier covers 500–5,000 devices. The docs state "there is no limit to the number of licensed machines that can concurrently download from a Connected Cache node." However, for sites with more than 5,000 devices, we recommend additional cache nodes based on scaling the largest hardware tier. This is a practical assumption — not an official limit — and should be validated through testing in your environment.
Per-Device Content Estimates Assumption
Microsoft documents which content types are cacheable but does not publish per-device monthly content volumes. The estimates used in this tool are approximations:
| Windows Updates (feature + quality) | ~500 MB/device/month |
| Microsoft 365 Apps (Office C2R) | ~300 MB/device/month |
| Intune Win32 / LOB Apps | ~200 MB/device/month |
| Microsoft Store Apps | ~50 MB/device/month |
| Windows Defender Definitions | ~150 MB/device/month |
| Autopilot Provisioning | ~4,000 MB/device (one-time) |
These are rough estimates. Actual volumes vary significantly by organization, update channel, and deployment configuration. They are used only to produce the "Estimated Monthly Client Demand" figure — the hardware sizing does not depend on these values.
Peak Delivery Analysis Field Observations
The Peak Delivery Analysis models what happens when many devices download content from the cache simultaneously. This calculation is based on field observations from real-world deployments:
- Cache throughput is shared: The cache node's sustained delivery rate (its NIC Gbps spec) is divided across all simultaneously downloading devices.
- Per-device bandwidth:
cache throughput (Kbps) ÷ simultaneous device count = per-device bandwidth (Kbps) - Download time:
content size per device ÷ per-device bandwidth = download time
Example: A site with 15,244 devices and a cache node delivering 5 Gbps. If all devices request 500 MB at the same time: 5,000,000 Kbps ÷ 15,244 = ~327 Kbps per device, giving a download time of ~3:24 per device.
Key mitigations: This is a worst-case scenario. In practice, configuring update rings to trigger updates at different times across the fleet significantly reduces simultaneous demand. Delivery Optimization peering (if enabled) further offloads the cache node by letting devices source content from each other. Use the "Simultaneous Download %" input to model more realistic scenarios.
Delivery Optimization Policies MS Learn
All DO policy names, values, and descriptions come from the Delivery Optimization reference. The cache server fallback delay of 60 seconds is a suggested starting point — Microsoft recommends setting this policy but does not specify a particular value. Adjust based on testing.
OS & General Requirements MS Learn
All operating system requirements, build numbers, WSL/Hyper-V dependencies, proxy warnings, port requirements, and NIC specifications come directly from the MCC prerequisites page.