20.09: March 2021 Update

Hello! We're excited to share the improvements in the March 2021 update for Backend.AI 20.09 / Enterprise R2! Version Information:

• Manager: 20.09.13
• Agent: 20.09.7
• Common: 20.09.7
• Storage-proxy: 20.09.1
• WebServer: 20.09.9
• WebUI: 21.03.1
• Control Panel: 20.09.7

Features for Administrators

Enhanced Statistics Viewing for Large-Scale Clusters

We've improved support for viewing usage statistics when many users access computing resources in large-scale clusters. You can now check node and session-specific usage history through the control panel interface with a more accurate and intuitive GUI. New Features:

• Resource usage history GUI by compute node
• Resource usage history GUI by session

Automatic User Creation and Deletion Feature

We've added a feature that simplifies the process of creating large numbers of temporary accounts for research and educational institutions. Simply specify ID rules and resource allocations through the GUI, and the system will automatically create the desired number of users using the template. When finished, accounts can be easily archived or completely deleted.

General Features for Users

FileBrowser Integration for Convenient File Management

In addition to Backend.AI's existing file management capabilities, we've integrated the FileBrowser project into Backend.AI's GUI. You can now access various file management features including:

• Drag-and-drop directory uploads
• File content preview
• Compressed downloads To use this feature, select the folder you want to manage from the Storage-Folder list in the Web or desktop app and choose "Open in FileBrowser."

Pre-setting Environment Variables

You can now specify environment variables to be used within a session before starting the session.

Development and Research Framework Support

Official Support for TensorFlow Distributed Auto-configuration

We now fully automate all user configuration requirements for distributed processing in TensorFlow 2.4 and above. When creating multi-node sessions, the remaining settings required for distributed processing are automatically configured based on the number of instances.

Official Support for Swift and Swift for TensorFlow

We officially support Swift REPL and Swift for TensorFlow, incorporating the Swift open-source build chain changes from October 2020.

Official Support for Julia and FluxML.jl

Backend.AI now officially supports Julia, the scientific computing language renowned for its speed and convenience. We've resolved various usability and security issues discovered during over two years of beta testing to provide a stable Julia environment. FluxML.jl support: We also officially support FluxML, the deep learning framework written in Julia. GPU acceleration support will be added in the next version.

R Studio Support (Beta)

In addition to the Jupyter-based R environment, we've added R Studio support in beta. You can select and use R Studio with R version 4 and above. Please share any issues or suggestions for improvement, and we'll continue to incorporate them before official support.

💡 March One-Line Tip

Did you know? You can directly connect to running sessions via terminal or copy files using backend.ai ssh and backend.ai scp from the CLI!