Developing the Custom Execution Engine

Our judge is built on a high-performance execution engine, integrated into our infrastructure to provide sub-second grading and absolute consistency across submissions.

• Sandboxed Execution: Every submission is executed in a fully isolated Docker (Alpine Linux) container. We enforce strict resource limits—specifically 1.0 CPU units and <256MB of RAM—to prevent denial-of-service attacks and ensure stability.
• Network Lockdown: All execution containers are started with the --network none flag. This ensures that student code has zero internet access, preventing any potential data leaks or unauthorized external communication.
• Read-Only File Systems: Containers are mounted with read-only filesystems, preventing malicious scripts from modifying the host environment or leaving persistent footprints.

• Custom Test Cases: We manually curated over 100+ test cases across 26 distinct problems in Sheet-1, covering edge cases from integer overflows to complex graph topologies. Our system provides detailed feedback for each case (WA, TLE, MLE, RE).
• Real-Time Progress: Submissions are graded asynchronously. As soon as a result is finalized, it's pushed to our global leaderboard via an optimized update pipeline, allowing students to track their progress relative to the cohort.
• Internal Data Tracking: Unlike public platforms, we track every attempt by every student. This allows our instructors to see exactly where students are struggling—whether it's on specific time limits or algorithmic complexity—and adjust our curriculum accordingly.

The execution system has transformed from a production-ready tool. By building our own execution layer, we've created a training environment that provides a precision instrument for academic growth. It scales seamlessly as we add new training sheets, providing a reliable platform for the next generation of developers at HUE.