Robocurse
Twenty years ago, a production system fell over. Millions of images on the site, hundreds of thousands in the replication queue. We had monitors on robocopy duration and file counts, so we watched it happening. Linear slowdown at first, understandable, more files means more time. Then it went non-linear. Fell on its face. Emergency replacement.
The fix cost us. Enterprise replication licensing. It worked, but most teams don’t have that budget.
Two and a half decades later, a client needed replication. Same constraints: Windows servers, locked files, large file counts. Different constraint: no enterprise budget.
What It Does
VSS-aware parallel replication for Windows, built to handle what robocopy alone cannot.
Core capabilities:
- Intelligent chunking: Distributes directories across workers based on size and file count, not just path depth
- Parallel execution: 1-128 concurrent robocopy instances, coordinated through a thread-safe orchestration layer
- VSS integration: Local snapshots and remote shadow copies via WinRM, so locked files don’t block replication
- Crash recovery: Checkpoint restoration when things go wrong mid-run
- Bandwidth throttling: Aggregate throttling across all parallel jobs, not per-job guessing
- Dual interfaces: Full WPF GUI for interactive use, headless mode for scheduled tasks
The Numbers
- ~1MB monolithic PowerShell script
- 35 public function files
- 47 unit + 9 integration + 10 enforcement test files
- 11 XAML templates, 13 GUI modules
- Production-tested in enterprise environments
Why Open Source
Built this for a client. Open-sourced it because nobody should have to go through robocopy pain or pay enterprise licensing for the privilege of reliable file replication.
The code uses Enforcement-Accelerated Development, which means the architectural rules are enforced by tests, not tribal knowledge. Fork it, extend it, break it differently.