What is Cloud-Native Video Analytics App?
As the name suggest, cloud-native video analytics apps are hosted in a private or public cloud. Cloud apps analyse thousands of video streams from city cameras and detects public safety threats in real-time, for example, weapon appearance, people laying down, or unattended luggage.
Cloud-native apps are built without employing a monolithic software codebase. Built in a modular manner and operating independently but on a common host and cloud operating system, cloud-native apps leverage cloud compute frameworks and infrastructures, which themselves comprise multiple, distinct, lightweight common services. With minimal overheads, cloud-native applications are flexible and highly efficient.
Is Edge Video Analytics Viable Alternative?
Modern smart cameras offer embedded video analytics and local storage, reducing or completely eliminating the need to process video in the cloud. This technology is designed to reduce the amount of data that needs to be transmitted from cameras to the datacenter. Edge video analytics is great for bandwidth-constraint deployments, e.g. on remote sites on a wireless connection.
In practice, in most public safety and smart city applications it is not possible to save bandwidth by storing and/or analysing video locally. In these applications dozens of users often need to quickly access the video storage, for example, in case of an emergency situation. Video is required to be stored off-site in case the site looses a network connection or destroyed. The other words, all video streams must be always transmitted to the central cloud in order to be accessed by many users at any time.
Another problem of edge analytics is a limitation of camera computational resources. Today, edge analytics do not achieve the same level of performance as cloud analytics with heavy apps such as suspect recognition or weapon detection on crowded scenes. At the same time, lighter apps such as people counting or intrusion detection can run at the edge without any accuracy compromise.
Why Cloud-Native Architecture is Optimal for Public Safety?
Sustainability: A city surveillance network has a very long lifecycle because of the camera and cabling installation costs. Over the last 10 years large cities like London and New York have deployed thousands cameras from multiple vendors. This infrastructure contains legacy hardware and can not be easily upgraded.
The cloud-native architecture decouples the video processor from the sensor, and further decouples the video analytics software from the hardware. Thus, all the layers of the city infrastructure can be upgraded independently, stage by stage. The cloud app can adapt to different types of sensors and video codecs without the costly need to replace cameras. Cameras are inexpensive and, if vandalized, can be replaced at the minimal expense.
High Availability: The cloud-native architecture is essential for mission-critical applications to enable 24/7 data analysis, recording, and response, even when the remote site is offline. Cloud-native apps provide no single point of failure and instant recovery in a more efficient manner than non-cloud apps running on virtual machines (VMs).
High Scalability: The cloud-native architecture enables cost-effective scalability across multiple metrics such as the number of cameras, the number of video analytics apps, the number of active users, the retention period, and suspect database size. By contrast, edge video analytics limits the number and the complexity of video analytics apps that can be applied for each video stream.
Is it Possible to Combine Cloud and Edge Analytics?
Yes, IREX video analytics modules can be deployed in a private cloud and edge servers, bringing the best of the two architectures. The cloud deployment is recommended for mission critical sites with a reliable network. The edge deployment works better for remote sites with bandwidth restrictions. Users can access the data from cloud and edge servers seamlessly, from in single web portal and mobiles apps.