Kubeflow Platform Managed Services

  • Enable your GPUs to be used effectively and efficiently
  • Subscription-based support, efficient upgrades, patches, and infrastructure monitoring
  • Leverage best practices from industry leading experts (Oreilly authors, Kubeflow contributors)
  • Quickly resolve critical infrastructure issues, producing less downtime and lower costs

The Patterson Consulting Managed Kubeflow offering provides an on-premise or hybrid (federated on-premise and cloud cluster) Kubeflow install customized to the enterprise's needs. This offering delivers a multi-tenant deep learning cluster to the organization in a turn-key fashion supporting production workloads with the most advanced hardware available. These multi-tennant clusters allow multiple teams of data scientists and data engineers to collaborate and share high-end compute and storage hardware. Our managed Kubeflow systems securely enable different execution modes such as distributed (TensorFlow, PyTorch, etc), multi-GPU deep learning, and distributed multi-GPU.

Platform management subscription includes:

Kubernetes Installation and Integration

  • Hardware Validation (DGX-1, Storage Arrays)
  • Networking Review and Architecture Design
  • Capacity planning based on usage profiles and analysis
  • Integration with Artifactory for On-premise container storage

Kubeflow Installation and Integration

  • Kubeflow installation and configuration
  • Multi-Tenancy configuration and customization
  • Gateway host configuration and integration
  • Customized job scheduling with active directory integration

Security Integration

  • Active Directory Integration
  • Kerberos Integration
  • Security Review

Flexible Hardware Profiles

  • Nvidia DGX-1 and DGX-2 (Nvidia documentation)
  • Cisco C480 ML
  • Heterogeneous mix of CPU, GPU hardware

Training, Documentation, and Support

  • Custom training classes delivered onsite for Kubernetes and Kubeflow
  • Custom internal run books for users
  • Ticketing system, Remote and on-site support
  • Customized monitoring of cluster operations

Integration with Existing Infrastructure

  • Custom integration with existing ETL pipelines
  • Custom integration with existing Active Directory systems
  • Integration with systems such as Cloudera, AWS, GCP, Azure, and more

Hybrid On-Premise / Cloud Kubeflow

We also offer hybrid Kubeflow clusters that can run machine learning workflows on an on-premise cluster and burst to a kubernetes cluster based on public cloud infrastructure as needed.

Contact Us for a Detailed Architecture Review

For more information on how we can help your organization, reach out to Patterson Consulting and we can work with your team to develop a customized Kubeflow cluster.