Managing Smart Surveillance Systems
The goal of a smart video surveillance systems must be the prevention of potentially catastrophic incidents, rather than an investigation of past incidents. The prevention would need constant monitoring of video data, and human monitoring of video data is very labor-intensive and almost economically inviable. So, an option will be to use computer vision. Let machines monitor the videos and identify mobility issues, criminal activities, production issues, manufacturing activities, and many more. The machine-assisted monitoring requires learning and machine learning requires data to learn, analyze and infer. In the context of video surveillance, the video data need to be collected, processed and analyzed. The quantum of data from video cameras is astonishing, an SD video camera generates around 3Mbps of data and 5 Mbps form an HD video. With simple math, an SD video generates 32 GB data in a day, which means just 10 cameras will generate 1.7 TB of data per month! A design pattern of processing this data in the cloud, in a data-center is very inefficient. It is best to process the video data on-site, on the edge of the network, close to the camera that is producing the data. Computing at Edge!
Why we need Edge Computing?
Smart surveillance is one of the edge computing applications. Edge Computing can also be used in applications that require autonomy (i.e. completing tasks with little or no human interaction) such as self-driving cars and Industry 4.0. Other areas where Edge Computing may find immense utility are applications that can’t tolerate latency. For example health care and financial transactions, where latency can be a reason for system failure.
The common issues in smart surveillance applications
Monitoring Hardware Failure
On-premise smart surveillance hardware includes among others, cameras, edge compute device, possible storage devices, a network device, and a compact power backup. The minimal solution includes at least a Local Processing Unit (LPU) and a camera that may be directly connected. Any system with more than one component has the probability of failure. We have come across cases where hardware failure due to on-premise power fluctuations, connectivity failure, and physical damage of devices. Hardware health must continuously be monitored and alert any failure to the right authority so that the issue can be fixed and the surveillance system comes back to normal operational status as early as possible. Any undetected failure can cause lax in the surveillance which can lead to failure in the prevention of catastrophic incidents.
Identifying camera outage
As it is said, a chain is as strong as its weakest link, any failures in the camera can cause the whole surveillance system to fail. Ability to monitor the camera outages, due to hardware failure, connectivity failure or physical damage, is very important to build a resilient smart surveillance system.
Distributed deployments
Smart surveillance systems that are deployed on-premise are geographically distributed. Unless systems and management solutions are built ground-up to support the distributed nature of the compute, the management of smart surveillance solutions can be very complex and may not allow scaling up to multiple sites.
Remote access
Reliable and Secure remote access mechanism is one of the critical ingredients for managing smart surveillance system. When the surveillance application is underperforming or not operational, the dev-ops team or operator should be able to connect to the edge computing device and debug the problem efficiently. Under many circumstances ability to debug a specific video stream from a specific camera may be essential, the remote access mechanism must support such functionality.
Monitoring Disk Space
As indicated above, the cameras can generate a huge amount of data and fill up the disk space very fast. The ability to manage and monitor the disk space becomes very critical for the normal operation of the smart surveillance system, failure to do so can make the system unusable. Management Solutions must alerts to the dev-ops team about the disk space unavailability. The system must be able to predict the usage pattern and provide an alert before a catastrophic failure can bring down the entire surveillance system.
Understanding Application outage
The edge computing surveillance application that process video data can be very complex. The application itself might include multiple components including computer vision, analytics, machine learning, etc. A failure of any one component of the application can cause an outage of the entire surveillance system. The ability to monitor individual components, identify bottlenecks and quickly get to the root cause of application issues is very essential to reduce the downtime.
OTA and Firmware Upgrade
While OTA (Over The Air) firmware and application upgrades are solved problems, it is an essential ingredient in smart surveillance solutions and needs a fresh look. The business agility requirements push for more frequent application upgrades. New learning algorithms or models for ML and AI applications call for frequent upgrades. Firmware (file system and kernel level) upgrades are also essential to plug security vulnerabilities.
Micro-data centers
In a large video surveillance systems that encompass an on-premise micro data center, the ability to understand the system and network issues, identifying performance bottlenecks, debug and fix the problem with minimal downtime is a must for the success of smart surveillance system.
While edge computing paradigm can be very beneficial in the development of a smart surveillance system, due to the nature of distributed architecture and remote location, the tools and techniques for performance management of application & infrastructure is an important piece of the smart surveillance jigsaw puzzle.
S Chetan Kumar is the co-founder and C.E.O of Aikaan Labs (www.aikaan.io), an Edge computing company. Aikaan labs help customer to do performance management of edge applications and infrastructure.
Chetan has close to 20 years of experience working in corporate and startup in the Internet and Systems Industry. Chetan has managed large engineering teams in the areas of Ethernet switching, security services, QoS and applications. Chetan has MSc degree in Electrical and Communication Engineering, from India Institute of Science and Bachelor degree in Electrical and Communication Engineering from Mysore University, India.
