In this special guest feature, Tim Miller, VP of Product Marketing at One Stop Systems (OSS), writes that his company is addressing the common requirements for video analytic applications with its AI on the Fly® building blocks. AI on the Fly is defined as moving datacenter levels of HPC and AI compute capabilities to the edge.
The increased efficiencies of GPU computing will also likely lead the path for edge computing. (Photo: Shutterstock/By YIUCHEUNG)
By 2020, over a billion video cameras used in commercial applications will be deployed, generating an avalanche of raw video data. In many cases, transforming this data to actionable intelligence requires deploying artificial intelligence and high performance computing capability at the data source. These diverse applications power smart city infrastructure, public safety and security systems, and drive business intelligence insights. In smart cities, video analytics is used for traffic management, parking optimization, crowd management, and infrastructure monitoring. Public safety and security applications at private and public sites including airports, sports arenas, and shopping malls are used for perimeter intrusion and anomaly detection, threat tracking, and importantly today, in mass health monitoring to identify potential virus hot spots when coupled with infrared thermometers. Business intelligence applications can drive real time retail strategies through dwell time and behavior analysis, and commercial operations management through real time event reporting and monitoring of industrial automation processes. Common functions in these applications include video data filtering, searching, aggregation, and image recognition. Artificial intelligence algorithms are often required to deliver transformative insights and enable real time decision making.
To enable these applications, the most powerful, high performance computing technologies historically associated with centralized enterprise and cloud data centers need to move out to the edge. The power required in these compute engines is increasingly being delivered through GPUs each with thousands of computational cores that can process massive data in parallel. Many applications can leverage NVIDIA’s DeepStream tools to harness GPU power for intelligent video analytics.
The common requirement for these video analytic applications is the ability to handle the ingress of large sets of video streams at a wide range of resolutions and bandwidths often aggregating across networks of hundreds or thousands of cameras. Although some data processing can be performed directly at the single camera, much of the analysis needs to be done in the edge area at aggregation points where real time decisions can be driven. Additionally, these solutions need to efficiently store this raw data once captured, and then move it to the third element of the compute and analysis engines. And uniquely because of the edge location, all of this capability needs to be designed and integrated to meet specialized size, power, and environmental constraints.
One Stop Systems (OSS) is addressing these requirements with its AI on the Fly® building blocks. AI on the Fly is defined as moving datacenter levels of HPC and AI compute capabilities to the edge. The three key elements of AI on the Fly solutions are data acquisition, data storage, and data compute. The strategy is to locate these solutions close to the data sources in the field, and convert that raw data to actionable intelligence without needing to move the data to remote or cloud-based datacenters. For OSS, these elements are delivered to the market as building blocks based on PCI Express with the ability to mix and match building blocks to address the needs of specific applications and deployments.
- Data acquisition systems: High PCIe slot count platforms to support a wide array of high-speed video acquisition cards enabling systems capable of acquiring hundreds of GB per second of data
- Storage systems: All flash servers and JBOFs for high speed, low latency data storage and data serving including NVMe storage array and JBOF nodes providing up to a petabyte of video data storage capacity
- Compute accelerators: Multi GPU/accelerator systems that are capable of processing massive amounts of data in parallel, including housing up to 16 of the latest NVIDIA GPUs or other HPC/AI accelerators for high end compute engine requirements
With OSS, all of these building block elements are connected seamlessly with memory mapped PCI Express interconnect configured and customized as appropriate, to meet the specific environmental requirements of ‘in the field’ installations.
OSS has the technology and expertise required to work with OEMs to build the next generation of AI on the Fly platforms which are currently being used by OEMs to build intelligent video analytic solutions deployed around the world today. Applications are emerging for this new paradigm not only in video analytics, but in diverse areas including autonomous vehicles, battlefield command and control, and industrial automation, genomic analysis, and location-based entertainment. As the increase in video data drives real time decision making, AI on the Fly delivers the technology OEMs require.
About the Author
Tim Miller is Vice President of Product Marketing at One Stop Systems. Tim has over 33 years of experience in high tech operations, management, marketing, business development, and sales. He previously was the CEO of Dolphin Interconnect Solutions and CEO and founder of StarGen, Inc. Tim holds a Bachelor of Science in Engineering from Cornell University, a Masters of Business Administration from Wharton, and a Masters in Computer Science from the University of Pennsylvania.
Disclaimer: This article may contain forward-looking statements based on One Stop Systems’ current expectations and assumptions regarding the company’s business and the performance of its products, the economy and other future conditions and forecasts of future events, circumstances and results.
Interesting post, Thanks for sharing.