BEAT Administrator Guide

BEAT framework is a project to build a platform for machine learning research and certification. By making use of such a system, academic or industrial parties enable users to easily compare results from distinct algorithms or/and parameterizations with minimal interaction.

This document contains resources for system administrators regarding the conception, dimensioning, hardware implementation and installation of a BEAT platform.

• 1. Overall Architecture
  • 1.1. Wedding List
• 2. Example: The Platform Deployed at Idiap Research Institute
  • 2.1. Deployment Strategy
  • 2.2. Hardware Specifications
    • 2.2.1. Computing Hardware
    • 2.2.2. Storage
    • 2.2.3. Summary
  • 2.3. Virtualization
  • 2.4. Software Specifications
  • 2.5. Storage Organization
• 3. Software Organization
  • 3.1. beat.backend.python
  • 3.2. beat.core
  • 3.3. beat.web
  • 3.4. beat.examples
  • 3.5. beat.cmdline
• 4. Deployment Guidelines
  • 4.1. Single machine
  • 4.2. Distributed processing nodes
  • 4.3. Load-balanced distributed architecture
• 5. Installation
  • 5.1. Installing beat.web
  • 5.2. Documentation
  • 5.3. Unit Testing
  • 5.4. End-to-End Testing
    • 5.4.1. Setup
    • 5.4.2. Running tests with the provided script
    • 5.4.3. Manual test running
      • 5.4.3.1. Starting the webdriver server
    • 5.4.4. Understanding the output of Protractor
      • 5.4.4.1. Saving test results
    • 5.4.5. Adding your test to Protractor
    • 5.4.6. Overriding Protractor’s browser choices
    • 5.4.7. Writing Protractor tests
      • 5.4.7.1. BEAT platform & Protractor’s Angular support
  • 5.5. Instantiating and Starting a Development System
    • 5.5.1. All-in-one Platform
    • 5.5.2. Discrete Platform using Localhost
    • 5.5.3. Mixing and matching
    • 5.5.4. Going full scale
  • 5.6. Development Notes
    • 5.6.1. Backup and Restore
    • 5.6.2. Testing Django Migrations
    • 5.6.3. Javascript Management with Node.js/Bower
  • 5.7. Issues
• 6. Validation Procedure for web interface
  • 6.1. Front page
  • 6.2. User page
  • 6.3. Search
  • 6.4. Experiments
  • 6.5. Toolchains
  • 6.6. Algorithms
  • 6.7. Databases
  • 6.8. Attestations
  • 6.9. Reports
• 7. Backend Maintenance
  • 7.1. Configuration
  • 7.2. Cache Maintenance
• 8. Extending the platform
  • 8.1. Not enough processing power
  • 8.2. Not enough room in the Data Cache
  • 8.3. The Web Server can’t process all the requests fast enough
  • 8.4. The Database Server can’t process all the requests fast enough