There are multiple options to choose from when designing an integrated computer vision (CV) and video management system (VMS) for a large environment. Most implementations include products from multiple vendors and then require a significant integration project effort before operations can begin. Dell Technologies together with three market-leading VMS and CV vendors have designed an integrated solution running on a Dell Validated Design for AI platform that is easy to operate and easy to grow. This document describes how we designed and tested our integrated solution on a common platform and presents some preliminary sizing guidelines that can jump-start an environment-specific design tailored to the needs of a large site.
This Design Guide describes how to implement virtualized instances of Milestone XProtect for video management together with CV applications from both BriefCam and Ipsotek. All three applications were implemented following guidance from the vendors on how to take advantage of native high availability (HA) features and scale-out performance that would be critical in an enterprise-class solution.
We chose the VxRail HCI system with VMware virtualization plus Nvidia A40 server-class GPUs. NVIDIA vGPU compatibility with the Dell Validated Design for AI platform provides better resource utilization and flexibility when implementing large systems with many virtual machines. The VxRail systems and VMware virtualization are design components that are already familiar to many IT and OT (operational technology) professionals. These components also provide additional features for managing high availability and scale-out needs that can complement the native features of the CV and VMS applications.
All VMS and Computer Vision products were installed on a single VxRail 5-node cluster and tested to demonstrate the value of a common platform. Storage for the video management archive functionality was provided by Dell PowerScale scale-out NAS storage. All video streams used for scale-out testing were provided by camera simulation software reading from prerecorded local video files in a continuous loop. The VMS high availability testing was performed with physical cameras since our simulation software did not support dynamic rerouting of output streams. We tested simulated camera streams being archived to the VMS simultaneously with two CV applications performing real-time analytics on a subset of the simulated camera streams.
This document confirms that the VxRail with VMware virtualization and NVIDIA GPUs provides a reliable and scalable Dell Validated Design for AI platform for hosting VMS and CV applications from different vendors on a common platform.