Supplementary Considerations for System Architecture
This article serves to provide details for our main article — Computer Aided Tablevision.
1. For IoT Things
- Managed/Unmanaged:
Managed IoT system allows control over contributors and coverage. The system is deployed at known locations (food centres) with fixed table setup. The continuous monitoring and calibration allow for better data quality. - Platform:
Low powered and standard Raspberry Pi. Standardised and available parts. Open sourced, license free and large community support and innovation. - Resiliency:
Decentralised processing avoids single point of failure. - Scalability:
Edge processing distributes the image processing requirements to the sensor nodes, which returns only the required data. It has better scalability compared to centralized processing, as typically a centralized server can only support up to 32 video streams and scaling beyond that is expensive. - Security:
Software deployment and updates, user account management using PiCockpit, key authentication for secure connection. - Power: Mains powered:
• Low maintenance as there is no need to change or charge battery
• Power is readily available / can be installed in the food centre
• Power multi-plexing by using n way USB power adapters instead of individual power adapters to minimize power point requirements. Slight reduction in resiliency but acceptable. - Deployment:
•Embedded device with camera sensor is placed on overhead beams/fixed structures in sheltered areas.
•No infrastructure change on tables
•Low IPx requirements (not affected by water, chemicals, weather)
•Less obstructions from human traffic
•May require period cleaning due to grease from cooking and dust from outdoor environment
2. For IoT Connectivity
- Performance requirements
Low data rate as transmitting processed data rather than sending image/ video streams. Short range (100m) as multiple access points are deployed under Wireless@SG. No power constraints as under mains supply. - Infrastructure & site constraints
Wifi is chosen due to the localised deployment with existing Wireless@SG coverage. Wifi can support high density deployment (typically 100 to 200 tables typical of a food centre, which is equivalent to number of sensor nodes required). - Topology: Flat network —
• high resiliency as individual gateway can fail without affecting others
• easy to scale by adding additional gateway corresponding to the number of tables
• infrastructure wireless mode provided under IEEE802.11 standard - Potential for interference
Possible interference from stall operators and customers running their own wireless LAN / hotspot for security cameras, POS and laptops. Wifi has auto channel selection, direct sequence spread spectrum (DSSS) and other technologies to avoid interference. - Security & regulatory needs
Dedicated SSID riding on Wireless@SG infrastructure provides better network security to prevent hacking. WPA2/WPA3 encryption to secure data transmission. - Indoor vs Outdoor
Indoor deployment. Stalls and tables have a fixed aisle and line of sight that allows good propagation of wireless signals. - Product Maturity
IEEE802.11 is well supported and easily available. As newer standards become available, there is an upgrade path available rather than having to switch technology.
3.For Splunk
In practice, a more elaborate setup involving load-balanced forwarders, clustered search heads and master nodes work together to deliver a highly resilient and scalable architecture to support distributed workloads. Such setups are usually deployed in internal secured networks with Virtual Private Network (VPN) access due to logs containing sensitive information that hackers are keen to access for information discovery or modify to cover up activity traces. For internet deployments, additional security controls such as Web Application Firewall (WAF), Intrusion Prevention System (IPS), Vulnerability Management, Security Monitoring, Backup & Recovery and Incident Response are required to safeguard against internet threats. Alternatively, all these infrastructure management can be outsourced through SaaS by subscribing to Splunk Cloud.
