ENTERPRISE AI ANALYSIS
Revolutionizing HPC Workflows with Drona Engine
Drona Workflow Engine is designed to simplify complex High-Performance Computing tasks for novice and experienced researchers alike. By providing an intuitive GUI, automated script generation, and a powerful framework, Drona minimizes the learning curve and maximizes research productivity.
Executive Impact at a Glance
Drona Workflow Engine delivers tangible benefits by abstracting HPC complexities and empowering researchers.
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Intuitive Workflow Execution
Drona provides a user-friendly Graphical User Interface (GUI) that guides researchers through the entire process of running scientific workflows on HPC resources. From selecting an environment to submitting a job, the GUI simplifies complex command-line interactions.
Key features include:
- Automated Script Generation: Drona creates all necessary batch scripts (e.g., Slurm) and configuration files based on user input.
- Editable Preview Window: Researchers can inspect and modify generated scripts with full syntax highlighting before submission, ensuring transparency and flexibility.
- Intelligent Feedback: The system offers real-time error, warning, and informational messages to assist users in providing correct and efficient input.
- Reproducibility: Every job context is stored, allowing for easy inspection of past runs and precise recreation or rerun of previous experiments.
Declarative Workflow Construction
Drona serves as a powerful framework enabling researchers and administrators to create, customize, and share scientific workflows in a largely declarative manner. This is achieved through a set of structured specification files:
schema.json: Defines the form elements (number, text, file selectors, dropdowns) that Drona renders in the GUI, including conditionals for dynamic display.map.json: Specifies how user input from the GUI is mapped to placeholders used within the workflow scripts and templates.driver.sh: The main script that orchestrates the workflow execution, which can involve submitting jobs to schedulers (Slurm, Kubernetes) or running complex pipelines.- Template Files: Reusable script fragments (e.g., Python, Slurm, config files) that Drona populates with mapped values.
utils.py(Support Library): Python functions for advanced input analysis, dynamic adjustments, and communication with the Drona Engine (e.g., adding messages, files, or mappings dynamically).
Enterprise Process Flow: Drona's Mapping Mechanism
Community & Scalability
Drona emphasizes ease of deployment and collaboration, ensuring workflows are accessible and manageable across various HPC environments and user groups.
Drona vs. Traditional HPC Job Composers
| Feature | Drona Workflow Engine | Traditional Job Composers (e.g., OOD, OpenComposer) |
|---|---|---|
| Interface | 100% GUI-driven, guided input forms | Mostly text-based or simple forms |
| Workflow Complexity | Supports complex, multi-job, dynamic pipelines via advanced mapping and driver scripts | Primarily variable substitution for single jobs | User Feedback | Advanced, real-time error/warning/note messages | Limited or no explicit feedback on input correctness |
| Workflow Creation & Sharing | Declarative framework for custom workflow creation, import, and sharing via Git repositories | Relies on pre-defined templates; less flexibility for custom development and sharing |
| Monitoring | Workflow-specific monitoring capabilities, beyond basic batch job status | Primarily basic batch job status (Slurm info) |
Drona is implemented with React for the frontend and Flask for the backend, and can be integrated into portals like OpenOnDemand. Environments are stored in external public GitHub repositories, facilitating community contributions and sharing.
Real-world Applications:
- IPU Tutorial: Used to simplify complex Graphcore IPU training tasks, abstracting Linux shell familiarity for participants.
- AI Dashboard: Consolidates AI/ML capabilities, providing streamlined access to diverse computational workflows for PyTorch, Hugging Face transformer models, and educational resources.
Quantify Your Savings with Drona
Estimate the potential time and cost savings for your organization by leveraging Drona's streamlined HPC workflow management.
Your Drona Implementation Roadmap
A structured approach to integrate Drona into your HPC environment and empower your research teams.
Phase 1: Discovery & Environment Setup
Assess current HPC workflows, identify key scientific applications, and set up Drona within your existing OpenOnDemand or standalone infrastructure. This includes configuring initial system environments.
Phase 2: Workflow Development & Customization
Collaborate with our experts to create or import Drona environments for your critical workflows. Customize schema.json and map.json files to match your specific input requirements and integrate with your existing batch schedulers.
Phase 3: User Training & Pilot Deployment
Conduct hands-on training for your research teams, focusing on Drona's intuitive GUI, script preview, and reproducibility features. Deploy Drona for a pilot group to gather feedback and refine workflows.
Phase 4: Full Integration & Ongoing Optimization
Roll out Drona across your organization. Leverage Drona's framework to add new form elements, integrate advanced monitoring, and continuously optimize workflows for maximum efficiency and researcher satisfaction.
Ready to Transform Your HPC Experience?
Connect with our specialists to explore how Drona Workflow Engine can streamline your scientific computing, boost productivity, and empower your research.