GPUManager Introduction

    Cubix GPUManager is software that provides management of GPU clusters in Cubix GPUXpander enclosures.
    The primary aims of GPUManager are asset management, health monitoring, and technical support data.
    

• Asset Management: GPUManager collects data from distributed GPU clusters into a central management console that
  • identifies all attached GPUXpander enclosures
  • identifies the GPUs within each enlosure
  • for each GPU, provides capabilities such as number of GPU cores, amount of RAM, etc.
  • identifies O/S and software driver version numbers
• Monitoring: GPUManager continuously monitors GPU clusters, and

  • for each enclosure, provides link status, number of lanes, and link speed
  • each GPU, provides temperature, fan speed, and GPU load characteristics

Architecture

GPUManager management console is a network application that communicates to GPU clusters using a software agent installed to the CPU in each GPU cluster.

The network connection between GPUManager and each CPU/GPU cluster uses TCP/IP on a Local Area Network and/or the Internet.