Skip to main content

Container Instance Overview

Container instances are the core service of the aifare platform, providing you with a powerful AI development environment. This page introduces the basic concepts, resource configuration, and directory structure of container instances.

What is a Container Instance

A container instance is a container that uses Docker technology for resource partitioning and isolation, offering less performance loss and higher efficiency compared to virtual machine instances.

Docker is not supported inside container instances. If you need to use Docker, please contact customer service to rent a bare metal server (bare metal servers are rented by the whole machine on a monthly basis).

Instance GPU/CPU/Memory Allocation

On aifare, you rent GPUs by selecting different instance types. The platform offers various instance types, such as:

  • Small Instance: 1 GPU + 8-core CPU + 32GB memory
  • Medium Instance: 2 GPUs + 16-core CPU + 64GB memory
  • Large Instance: 4 GPUs + 32-core CPU + 128GB memory
  • Extra Large Instance: 8 GPUs + 64-core CPU + 256GB memory

The platform creates instances based on instance types, which are flexibly designed to combine different sizes of GPU, CPU, and memory. You can choose the appropriate instance type according to your project needs:

  • Learning and Small Projects: Choose small instances for lower cost
  • Medium-scale Training: Choose medium instances for a balance of performance and cost
  • Large Model Training: Choose large instances for sufficient computing power
  • Ultra-large-scale Training: Choose extra large instances for multi-GPU parallel training

Each instance type is optimized to ensure the best match of GPU, CPU, and memory resources, providing a stable and efficient environment for your AI development.

Directories in the Instance

To check disk usage in the instance, run in the terminal: source /root/.bashrc

NamePath (except special paths below)SizePerformanceDescription
System DiskRoot / and all subpaths30GBLocal, fastData is retained after shutdown. System dependencies and Python packages are usually installed here, as well as code and small data. Migrated with the instance and saved in custom images.
Data Disk/data30GB+Local, fastData is retained after shutdown. Suitable for high I/O data. Not saved in images. For data disk migration, see instance migration (same region).
Model Storage/gm-models1.0TBNetwork, normalPrebuilt model storage with common AI models for quick access and use.
Dataset Storage/gm-datasets1.0TBNetwork, normalPublic dataset storage with common training datasets for quick loading.
User Data Storage/user-data1.0TBNetwork, normalPersonal data storage, supports cross-instance data sync and sharing.

More on the System Disk

In Linux, almost everything is a file or folder, all under the root directory /. If you install Linux on a disk A (with no other partitions), the root directory will have default system folders like:

$ ls /
bin etc lib mnt proc run tmp var boot dev home root srv sys usr

JupyterLab Working Directory

The working directory of JupyterLab is /root. Data disks, network disks, and public data directories are all under /root, so you can directly see these platform-created directories after entering JupyterLab. To view the root directory /, use the terminal and command line.

Check Disk Usage

Run the following command in the terminal to check detailed disk usage:

df -hT

Directory Description:

DirectoryNameSpeedDescription
/System DiskNormalData is retained after shutdown, can store code, saved with custom images.
/dataData DiskFastData is retained after shutdown, suitable for high I/O data, not saved with images.

Notes:

  1. For system disk cleanup, see: https://www.aifare.com/docs/qa/

Instance Status

Container instances have the following statuses:

  • Running: The instance is running and ready for development
  • Stopped: The instance is stopped, but data is retained
  • Creating: The instance is being created and initialized
  • Released: The instance has been deleted and data is cleared

Main Features

Core Features

  • Overview: View basic information, resource configuration, and running status
  • JupyterLab: Online development environment supporting Notebooks and code editing
  • SSH: Connect to the instance via SSH for command line operations

Configuration Management

  • Scale Up/Down: Adjust CPU, memory, GPU resources
  • Switch Billing Mode: Switch between pay-as-you-go and monthly billing
  • Save Image: Save the current environment as a custom image

System Operations

  • Reset System: Restore the instance to its initial state
  • Migrate Instance: Migrate the instance to another node or region
  • Migrate Instance (Same Region): Migrate the instance within the same region

Usage Suggestions

  1. Choose the right configuration: Select CPU, memory, and GPU based on project needs
  2. Use storage wisely: Store code on the system disk, large data on the data disk
  3. Save data in time: Upload important data to persistent storage promptly
  4. Use resources efficiently: Shut down when not in use to save costs
  5. Back up important environments: Regularly save images to back up your development environment

Next Steps