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
Enhancing the Dell Validated Design for Computer Vision with BOSS AI on Tanzu Kubernetes Grid
Introduction Solution overview Partner overview Solution architecture
Solution architecture
Introduction
Enabling TKG on the CV platform
BOSS AI on TKG
Accessing Ipsotek Data from BOSS AI
Deployment and support
Conclusion

BOSS AI on TKG

Home > Workload Solutions > Computer Vision > White Papers > Enhancing the Dell Validated Design for Computer Vision with BOSS AI on Tanzu Kubernetes Grid > BOSS AI on TKG

BOSS AI on TKG

  • BOSS AI Platform On-Premises Deployment Requirements

    The BOSS AI platform can be deployed onto any size of TKG cluster. The most commonly deployed cluster configurations are listed below:

    Cluster category Number of workers

    (physical servers

    		CPU RAM (GB) Local storage (GB)
    Small 3 8 (16 Cores Min) 32 300
    Medium 6 16 (16 Cores Min) 64 300
    Large 9 24 (16 Cores Min) 96 300

    predeployment requirements

    • Load Balancer
    • Persistent Storage
      • Backed by a resilient fault-tolerant datastore.
    Note: Before deployment begins, any exceptions should be discussed with the BOSS AI support team .

    Load Balancer

    Every pod on a cluster requires a unique cluster-wide IP and a suitable method for pods to communicate with the outside world.

    BOSS AI requires that the following TCP Ports are free and open for cluster ingress:

    • 80, 443
    • 5671 (If used in a federated environment)
    Note: VMware TKG offers multicloud load balancing, a web application firewall, and an NSX ALB  container ingress service. Other load balancers can also be used with BOSS AI.

    Persistent Storage

    BOSS AI recommends a minimum of 1 TB of persistent storage availability before deployment. The Dell VxRail cluster was configured with vSAN datastores available to TKG as persistent storage.

    Note:
    • Storage requirements are use-case dependent.

    • Dell VxRail includes a Container Storage Integration (C.S.I) interface to enable persistent storage consumption with VMware TKG. Other storage options are also supported, such as PowerScale.



    BOSS AI Architecture

    The following figure shows a high-level overview of the BOSS AI system architecture:

    Figure 2. BOSS AI system architecture

    Summary

    BOSS AI Enterprise is deployed as a container-based architecture.

    Core services such as security, load balancing, data ingestion, and monitoring are deployed as container pods, enabling cloud and on-premises deployments.

    Common AI functions such as Extract, Transform, and Load capabilities are exposed to the AI/ML practitioner, with low or "No Code" options available.

    • Several data ingestion options are available, such as local file uploads or automated workflows provided by Apache NIFI.
    • Traditional data querying can be done using the Elasticsearch domain-specific language (DSL) syntax.
    • Simplified visual querying can be done with drag-and-drop interfaces for less technical users.
    • Data transformation is supported using an intuitive drag-and-drop "transformation tree" building in the BOSS AI GUI.
    • Feature selection is simplified with BOSS's no-code solution automatically visualizes feature selection, which applies throughout the entire ML development workflow.

    For more details on BOSS AI features, services, and integrations, see BOSS AI Documentation.

    BOSS AI deployment on TKG

    BOSS AI Enterprise is deployed as containerized pods to worker nodes within the TKG cluster on the CV platform.

    Figure 3. BOSS AI deployment on TKG overview

    The addition of container support provides the ability to manage High availability and scale-out requirements of CV and VMS applications for both VM and container-based workloads on VxRail solutions.