Scaling Strategies (Vertical & Horizontal Architecture)

As video surveillance installations grow from dozens to thousands of camera streams, RUSEON Core supports both Vertical Scaling (Scale-Up) on high-density bare-metal hardware and Horizontal Scaling (Scale-Out) across distributed edge gateway clusters.

---

1. Vertical Scaling (Scale-Up) Best Practices

A single modern 12-to-16 core server can handle 600+ concurrent 1080p streams (18,000+ FPS) when optimized:

1. NUMA Binding with numactl: On dual-socket servers, bind the process to a single NUMA node to prevent cross-socket memory bus latency:

   numactl --cpunodebind=0 --membind=0 /opt/ruseon/bin/ruseon -config /etc/ruseon/config.yaml

2. Dedicated Storage Tiering:
   • BadgerDB (./data): Place on dedicated NVMe SSD for microsecond LSM compaction.
   • Video Archive (./recordings): Place on large sequential arrays (ZFS RAID-Z2 or hardware RAID 6).
3. Socket Receive Buffer Tuning: Set net.core.rmem_max=33554432 in /etc/sysctl.conf to buffer network burst traffic.

---

2. Horizontal Scaling (Scale-Out) & Camera Sharding

For multi-building campuses, distributed retail chains, or smart city deployments exceeding 600 cameras:

• Camera Sharding: Assign camera groups to independent edge nodes based on physical location (e.g. Node 1 = Building A, Node 2 = Building B).
• Unified Ingress Proxy: Place an Nginx or HAProxy reverse proxy in front of the nodes, routing camera WHEP and HLS requests by camera ID prefix:

  location ~ ^/(stream|api)/(hls|webrtc|cameras)/(cam_bldgA_.*) {
      proxy_pass http://edge-node-1.internal:8080;
  }
  location ~ ^/(stream|api)/(hls|webrtc|cameras)/(cam_bldgB_.*) {
      proxy_pass http://edge-node-2.internal:8080;
  }

Related Documentation

• Performance Overview
• Real-World Benchmarks
• Capacity Planning Guide
• Reverse Proxy Setup