CKM (Cloud Kernel Manager)
A workload orchestration system built to understand how kernel scheduling, memory management, and production infrastructure come together.
CKM
Overview
CKM started as a scheduler simulation and evolved into a system that executes real Docker containers with production-grade patterns. It implements various scheduling algorithms (FIFO, Round-Robin, Fair, Priority) and connects them to real-time metrics, bridging the gap between OS theory and cloud reality.
Key Features
- Multiple Schedulers: FIFO, Round-Robin, Fair, Priority, Multilevel
- Resilience Patterns: Circuit Breaker, Rate Limiter, Graceful Shutdown
- Real-time Observability: Prometheus metrics and Grafana dashboards
- Container Discovery: Automatically monitors all running Docker containers
- REST API: Full CRUD for workloads
What I Learned
- ✨ Scheduling isn't just an algorithm; the right choice depends heavily on workload type (interactive vs batch).
- ✨ Metrics are essential storytelling tools for system behavior, not just numbers.
- ✨ Production systems are about handling failure gracefully (Circuit Breakers, Rate Limits).
- ✨ Gap between CS theory and production systems is smaller than expected.
Tools & Technologies
Go (Golang)
Docker Runtime
Prometheus
Grafana