Your Browser is Out of Date

Nytro.ai uses technology that works best in other browsers.
For a full experience use one of the browsers below
Computer Vision for Public Transportation - Phase 2
Overview Architecture concepts Solution architecture
Solution architecture
Physical architecture
Computing resources Network Storage resources
Virtual architecture
Bus depot Data center private cloud EPIC iO managed cloud
Software
Validation Sizing the solution Conclusions References

Physical architecture

Home > Workload Solutions > Computer Vision > Guides > Computer Vision for Public Transportation - Phase 2 > Physical architecture

Physical architecture

  • Computing resources

    Bus compute

    We partnered with EPIC iO systems to deploy a Deep Insights edge device in our lab environment. This solution is made up of compute, networking, and power that can be easily deployed for any transportation use case.

    The specification of the on bus compute capability is:

    Model
    Dell OptiPlex 7090 (Micro Form Factor)
    CPU
    Intel(R)Core(TM) i9-10900T CPU @ 1.90GHz
    Memory
    32 GB
    Storage
    2 TB NVMe

    This compute node is responsible for running the CV analytics on the bus in real-time to ensure that the bus driver can be made aware of any alerts in while out on the bus route.

    The prebuilt EPIC iO appliance also contains a ruggedized dual sim 5G Cradlepoint router. This allows seamless communication with the bus while out in the field over public 5G and more importantly it also contains a sim card that allows the bus to communicate over a private 5G network.

    Depot compute

    A bus depot (or any depot) does not typically contain an enterprise-grade data center to host IT infrastructure. To account for this, we have designed the bus depot so that it can use the ruggedized Dell PowerEdge XR4000 platform. The XR4000 platform has the following benefits:

    • Withstands temperatures in the range -5 to +55 celsius (27F - 131F)
    • Handles dust, high humidity, and airborne contaminants
    • Can withstand shocks and vibration on factory floors or mobile platforms
    • Has multiple form factors and can contain between 1 - 4 compute servers (sleds)
    • Has wall mountable and rack mountable options to cater for any environment

    For our testing the following XR4000 cluster was configured:

    Model
    PowerEdge XR4000r (2U Rack mountable)
    Sleds
    2 x XR4520c Compute Server, 1 x XR4000x Witness Node
    CPU
    16 CPU Cores/sled
    Memory
    128 GB/sled
    Storage
    20 TB vSAN storage
    Note: vSAN storage was configured at the depot to provide shared storage across the servers in the XR4000 cluster. This is important for resiliency at the depot.

    Data center compute

    Our lab CV platform consists of a VxRail V670F 5-node cluster that is connected to an A3000PowerScale.VxRail is a hyperconverged infrastructure (HCI) appliance that is jointly engineered by Dell Technologies and VMware. VxRail combines computing, storage, and networking components in a single, compact unit to simplify the deployment and management of virtualized workloads.

    In most deployments the CV Platform will host customers VMS and CV workloads that are distributed across the VxRail cluster. Additional workloads, such as ProHawk video restoration and Seoul Robotics LiDAR analytics can also run on the same cluster.

    An example of a 5-node cluster is below:

    Table 1. Cluster components
    Resource Per Node Total for Cluster
    CPU 2 x Intel(R) Xeon(R) Gold 6354 CPU @ 3.00 GHz (72 vCPU) 724 GHz (360 vCPU)
    Memory 512 GB 2.5 TB
    Storage (vSAN) 27 TB 135 TB
    GPU 2 x A16 64 GB NVIDIA GPU 10 x A16 64 GB NVIDIA GPU
    Note: This solution scales horizontally by adding additional nodes when needed.