One Convergence Highlights DKube Deep Learning-as-a-Service Application at KubeCon 2018

  • On-Prem Performance with Cloud Native Simplicity and Mobility
  • Model experimentation, training and deployment with intuitive, GUI-based platform
  • Install and be working on models within 4 hours
  • Integrated package that enables use of standard platforms - all working out-of-the-box

KubeCon 2018 – Seattle, WA – November 11, 2018 – One Convergence, the software infrastructure company for cloud native data centers, will be showcasing the DKube on-prem Deep Learning application at this year’s KubeCon conference.

DKube allows organizations to train and deploy their sensitive, resource-intensive deep learning models on their own servers. It provides cloud native capability, but offers this simplicity and scalability on secure local machines or dedicated clusters provided by bare metal cloud providers.

The DKube application offers features that are normally associated with public cloud applications, but have not previously been available for organizations that preferred - or been required - to keep the models and data on their on systems.

  • On-demand GPU allocation - GPUs are allocated to users and jobs from a flexible pool, and are available for use whenever they are idle. This puts the resources where they are needed, but allows for the best utilization of scarce, expensive devices.
  • Distributed composability - Cluster-wide resources are available to any node on the cluster. This allows the resources to be allocated where they are required regardless of physical location. It enables rapid and simple scale-up or scale-out, with new devices automatically recognized and put to use. RDMA provides high bandwidth and low latency communication the node on a cluster.
  • Collaboration - Users can share models, datasets, and resources with enterprise-grade security. Users are authenticated before being allowed onto the system, and can only access what they are authorized for. Groups of users can co-exist, and be working with models simultaneously.

But these are just table stakes. DKube offers advantages beyond the basic platform capabilities.

  • Users can install and be working on models within 4 hours.
  • The DKube product is based on open platforms. It will run out-of-the-box on standard hardware systems. And it integrates the best-in-class standard frameworks such as Kubeflow, Jupyter, TensorBoard, and PyTorch. New frameworks can be rapidly added as required. DKube brings together the standard hardware and software components, and ensures that they all run together seamlessly.
  • DKube facilitates its advantages with an intuitive, GUI-based workflow.

One Convergence will be showing the latest version of DKube at the conference, demonstrating the UI-based workflow that enables users to be effective quickly.

About One Convergence

One Convergence, Inc. provides infrastructure solutions to optimize the cloud native data center. The company focuses on delivering simplified, secure, scalable software to drive enterprise customers and Cloud Service Providers. One Convergence has proven experience in networking, security, cloud native infrastructure, deep learning workflows, and server interconnect fabrics, and applying them to build the distributed software required for modern enterprise platforms. For more information, visit www.oneconvergence.com.