Sizing Your AiVR

Every AiVR has two limits: how many camera streams it can take, and how much video it can decode for AI. A design must pass both, plus the site's upload speed and the local disk. This article explains how to count streams, calculate decoding units, choose an AiVR model or stack, and check upload and storage before you quote.

 Note: The figures in this article are planning estimates for sizing and quoting. Actual capacity, bandwidth and storage use depend on each camera’s settings, scene activity, firmware and network conditions. Busy scenes, night IR and rain raise bitrates; quiet scenes lower them. Always validate a design against the real camera settings on site, and review the AiVR’s Bandwidth and Storage tabs after installation. For sites with more than 50 streams, multiple networks, or unusual camera types, contact your Alocity account manager or support@alocity.com to confirm the design before you quote. 

1. The Four Checks

The stream limit applies even when units remain, and units apply even when stream slots remain.

2. Count Streams, Not Cameras

Alocity sizes and bills by stream. The stream the AiVR records is the same stream the AI analyzes, so there is no separate inference stream to count.

Note: A fisheye that exposes a fisheye channel and a quad channel is two streams if you add both. Pull only the channel the site needs.

3. Calculate Decoding Units

1 unit = one 1080p stream at 15 fps. Units scale with pixel count and frame rate:

Units = (width × height ÷ 1920 × 1080) × (fps ÷ 15) × AI role multiplier

Then multiply by the camera's AI role:

Note: To adjust camera streams please see this help article: Best Camera Settings

Quick lookup: units per stream

Find the stream's resolution and frame rate, then read across to its AI role. Multiply by the number of streams.

IMPORTANT: One 4K camera at 30 fps uses as much capacity as eight 1080p cameras at 15 fps, which is half an AiVR-16. Resolution above 1080p adds little to basic detection, because each analyzed frame is a 640×640 copy. Get detail by narrowing the field of view, not by raising resolution.

4. Choose the Model

Note: Unit budgets already keep about 30% headroom for motion bursts. Pick the smallest model that passes both streams and units.  Every site is different. This is all to be used as reference only.

Stacking AiVRs

• For bigger sites, combine AiVRs. Models can be mixed.
• Each stream runs on one AiVR, and each AiVR must be on the same network as its cameras. Split cameras by network first, then size each group.
• Prefer the fewest boxes. At 50 or more streams, an AiVR-60 usually makes more sense than several smaller units.

5. Worked Examples

To fix example C, move two of the 30 fps cameras to 15 fps. That brings it to 16 units, within budget. Alternatively, move to an AiVR-60 or stack a second AiVR.

6. Check Upload

The AiVR backs up video to the cloud continuously. Site upload must be at least the sum of all camera average bitrates plus 20%, sustained.  See: Network Requirements

• Example: 10 cameras × 4 Mbps = 40 Mbps, plus 20% ≈ 48 Mbps.
• With less upload, video still records locally, but cloud backup falls behind.
• Use average bitrates, not maximums. Starting averages for the Efficient H.265 profile at 15 fps, See: Best Camera Configurations for AiVR.

7. Check Disk

1 Mbps of average bitrate ≈ 10.8 GB per day ≈ 324 GB over 30 days. Add up the average bitrate of every stream to get the daily ingest, then compare it with the AiVR's drive or NAS.

• Example: 10 streams × 2 Mbps = 20 Mbps ≈ 216 GB per day, so a 1 TB drive holds about 4–5 days on site.
• Footage older than local retention is served from the cloud, within the cloud retention window.
• If the customer needs more days on site, add a NAS. Use surveillance-class CMR hard drives or SSDs only. Consumer SMR drives cause recording gaps on every camera at once.

After installation, check real ingest on the AiVR's Storage tab (DAILY INGEST and DAYS OF VIDEO IN LOCAL STORAGE). See Getting to Know Your AiVR.

8. Sizing Checklist

1. List every camera with its type, sensors, resolution, frame rate and AI role.
2. Count streams.
3. Calculate units, including the AI role multiplier.
4. Pick the best fit model, or stack, that passes streams and units.
5. Check upload against the site's measured upload speed.
6. Check local disk against the days the customer wants on site.
7. Budget labor for configuring the Alocity stream on every camera. The AiVR can't change camera settings, so each camera is set up by hand for the time being. 

For questions or assistance, please contact your account manager or email support@alocity.com.