Clusters

Tensor One clusters are containerised instances intended for application execution.

Note: To guarantee compatibility across supported architectures, use the following build flag when creating images for Tensor One:

--platform Linux/Arm64, Linux/AMD64

Comprehending Cluster Configuration and Components

You can create a cluster, which is a containerised server instance, to make use of particular hardware resources. A distinct, dynamically generated identifier (such as 2s56cp0pof1rmt) is assigned to each Cluster.

A typical cluster consists of:

• Container Volume:

  • Holds the operating system and temporary storage.
  • This storage is volatile and is lost upon Cluster shutdown or reboot.

• Disk Volume:

  • Provides permanent storage, preserved throughout the Cluster’s lifespan.
  • Storage remains available even after a Cluster restart or shutdown.

• Ubuntu Linux Container:

  • Runs any software compatible with Ubuntu Linux.

• Resource Allocation:

  • Dedicated vCPU and RAM resources for the container and its processes.

• Optional GPUs or CPUs:

  • Specialized resources for workloads such as CUDA, AI, or machine learning tasks.

• Pre-configured Templates:

  • Automatically installs software and settings upon Cluster creation.
  • Enables streamlined, one-click access to common tools and packages.

• Proxy Connection for Web Access:

  • Provides connectivity via proxy URLs.
  • Format example: https://[Cluster-id]-[port-number].proxy.tpo.one, e.g., https://2s56cp0pof1rmt-7860.proxy.tpo.one.

Customizing Your Cluster

For further customization, you can configure the following options:

• GPU Type and Quantity
• System Disk Size
• Start Commands
• Environment Variables
• HTTP/TCP Port Exposure
• Persistent Storage Options

To begin, follow instructions on how to Choosing a Cluster and Managing Clusters.