News

When planning a video surveillance system, camera count is important—but it is not the only factor.

A 32-channel system recording standard 1080p video requires far fewer resources than a 32-channel system processing 4K video with multiple AI analytics applications.

If the server is undersized, users may experience video delays, slow playback, insufficient storage, or reduced AI performance. If it is oversized, the project may cost more than necessary.

The right system should be selected based on:

• Camera count and resolution
• Video retention period
• Number of simultaneous AI analytics channels
• Storage and network capacity
• Redundancy and future expansion

AI VMS Mini and AI VMS Server are powered by Argo AI VMS. Both support camera management, live viewing, video recording, and AI analytics. The main differences are system capacity, AI performance, storage, and expandability.

---

Five Questions to Ask First

Before choosing a server, confirm:

1. How many cameras will you manage now and in the future?
2. What resolution and bitrate will the cameras use?
3. How many days of video must be stored?
4. How many channels need to run AI analytics simultaneously?
5. Do you need redundancy or multi-site management?

For a small site with fewer cameras and basic AI applications, AI VMS Mini may be sufficient. Larger sites with higher recording, AI processing, or redundancy requirements may need an AI VMS Server or a distributed system.

---

AI VMS Mini: For 16–32 Channels

AI VMS Mini is suitable for small or distributed sites, including:

• Retail stores
• Construction sites
• Offices
• Clinics
• Residential buildings
• Warehouses
• Parking lots
• Branch offices

Its compact design and low power consumption make it easy to deploy. It can manage live video, local recording, and AI event detection at each site.

For a retail chain or multi-branch company, an AI VMS Mini can be installed at each location for local recording and AI processing. Headquarters can then centrally monitor system status, important events, and selected video.

If the external network connection is interrupted, each site can continue recording locally.

---

AI VMS Server: For 64–256 Channels

AI VMS Server is designed for medium to large sites with more cameras, longer retention periods, or heavier AI workloads.

Typical applications include:

• Factories and large warehouses
• Shopping centers
• Schools and hospitals
• Corporate campuses
• Transportation hubs
• Public safety applications
• Critical infrastructure

2U Server

The 2U model provides a balance of computing power, storage capacity, and rack space. It is suitable for most medium to large surveillance projects.

3U Server

The 3U model provides more space for hard drives, GPUs, NPUs, and future expansion. It is suitable for heavy AI workloads, longer retention periods, and large storage requirements.

The choice between a 2U and 3U server should not be based on camera count alone. A 128-channel recording system may fit a 2U server, while a 128-channel system running multiple AI models may require a 3U server.

For larger projects, recording, AI processing, storage, and failover can also be distributed across multiple servers to improve performance and reliability.

---

AI VMS Mini vs. AI VMS Server

Item	AI VMS Mini	Server 2U	Server 3U
Recommended size	16–32 channels	64–256 channels	128–256+ channels
Best for	Stores, offices, construction sites, clinics	Factories, malls, schools, hospitals	Transportation, large campuses, critical infrastructure
AI workload	Light or specific AI tasks	Multiple AI channels	Multiple models, VLM, and heavy AI workloads
Storage	Built-in or external storage	Medium to large RAID storage	Greater storage and expansion capacity
Main benefit	Compact and easy to deploy	Balanced performance and capacity	Maximum AI, storage, and expansion

These channel ranges are general guidelines. Actual capacity depends on resolution, bitrate, recording settings, AI models, and the number of simultaneous users.

---

Four Key Selection Factors

1. Video Workload

Do not look only at camera count. Also check:

• Resolution and frame rate
• H.264 or H.265 compression
• Average and peak bitrate
• Main stream and substream use
• Number of live-view users
• Playback and video export requirements

For example, 32 high-bitrate 4K streams may require more resources than a larger number of 1080p streams.

Using substreams for multi-camera viewing and main streams for full-screen viewing or recording can reduce network and playback loads.

2. Storage Capacity

Storage requirements depend on:

• Number of cameras
• Video resolution and bitrate
• Daily recording hours
• Retention period

A system storing video for 90 days needs much more space than one storing video for 30 days.

You should also reserve capacity for:

• AI event clips and thumbnails
• System files and databases
• RAID protection
• Future cameras
• Bitrate changes

Spark can help estimate the required storage capacity based on your camera settings and retention policy.

3. AI Processing

A system may support many AI functions, but that does not mean it can run all of them on every channel simultaneously.

Confirm the following:

• Number of AI channels running simultaneously
• Required analytics frame rate
• Main-stream or substream analysis
• Number of AI models per channel
• Facial, license plate, or detailed object recognition requirements
• Cross-camera tracking
• VLM indexing and semantic search

Different AI applications require different levels of computing power. People counting usually requires fewer resources than facial recognition, license plate recognition, or VLM-based search.

AI analytics can also be scheduled or triggered by events instead of running continuously. This helps balance accuracy, processing speed, and cost.

4. Network and Deployment

For a single site, make sure the switches and network backbone can handle all video streams.

For multiple sites, also consider:

• Upload bandwidth at each location
• Number of streams viewed at headquarters
• Local recording during network outages
• Full video or event-only transmission
• Remote user connections
• VPN or secure network requirements

A distributed system processes video locally and sends only alerts, thumbnails, metadata, and selected video to headquarters. This reduces bandwidth use and is ideal for retail chains, factories, and multi-site businesses.

---

Redundancy and Future Expansion

For factories, hospitals, transportation systems, and critical infrastructure, the system must continue operating when equipment or network problems occur.

Important options include:

• RAID storage protection
• Server failover
• Device health monitoring
• UPS backup power
• Local recording during network outages
• System audit logs

Future expansion should also be considered. You may need to add cameras, extend retention periods, deploy additional AI models, or integrate access control, building automation, POS, ERP, and intelligent operations systems.

---

Five Common Mistakes

• Choosing a server based only on camera count
• Underestimating storage requirements
• Assuming every AI function can run on every channel
• Using all system capacity from day one
• Ignoring redundancy and network outages

---

How Spark Helps You Choose

Spark provides flexible options for different project sizes:

• AI VMS Mini: For small or distributed sites with 16–32 channels
• AI VMS Server 2U: For medium to large video and AI projects
• AI VMS Server 3U: For heavy AI workloads and large storage requirements
• Argo AI VMS: For centralized management of video, devices, users, and AI events

Our recommended selection process is:

1. Check camera count and video settings.
2. Calculate storage requirements.
3. Confirm AI models and concurrent AI channels.
4. Plan the network and multi-site architecture.
5. Review redundancy and future expansion.
6. Test the system with real project video.

The best server is not always the largest one. It is the system that provides the right balance of performance, reliability, cost, and future growth.

---

FAQ

Is AI VMS Mini always suitable for 32 cameras?

No. If all 32 cameras use high-resolution video and run multiple AI analytics, an AI VMS Server may be required.

Should I choose a 2U or 3U server?

Choose 2U for standard medium to large projects. Choose 3U when you need more AI processing, storage, or expansion capacity.

Can one server manage 256 cameras?

It depends on the recording, AI, storage, and redundancy requirements. A distributed system is often more reliable for large projects.

Does AI always require high-resolution video?

No. People counting and virtual fencing may work with lower-resolution streams. Face, license plate, and detailed object recognition usually require more image detail.

Must all video be sent to headquarters?

No. Each site can record and process video locally, then send only events and selected video to headquarters.