The price of running Fkh (Freddy’s Kubernetes Helper)
Fkh runs in your own Azure subscription. This post lists every cost component, explains how each one scales with concurrent containers and Windows node size, and gives example monthly bills. With an 8-hour working day and scale-to-zero, cost per container falls to roughly $30 to $60 at 10 to 40 containers. Running 24/7 costs several times more, so StopFkh, spot nodes and right-sizing matter.
post Freddys blog
"The per-container cost drops as you grow, because the fixed baseline is shared across all running containers, and larger nodes pack more densely."
Key points
- Four cost buckets exist: always-billed infrastructure, usage-based scaling, conditional features, and free components
- Windows node pool compute dominates variable costs and scales to zero outside working hours, while Linux SQL node and Premium SSD form a fixed baseline
- Per-container cost decreases as concurrency increases because fixed baseline spreads across more workloads; larger node SKUs pack containers more efficiently
- Scaling to zero during non-work hours, right-sizing nodes, and using spot instances are the main levers to control costs
- Linux-based Business Central containers would cut compute costs roughly in half when they become available
Quotes
- “The per-container cost drops as you grow, because the fixed baseline is shared across all running containers, and larger nodes pack more densely.” (Explains the key cost-efficiency advantage of scaling: fixed costs amortize, making larger deployments cheaper per unit)
Context
- Features: Windows node autoscaling, Load Balancer per-container rules, Premium SSD persistent storage, SQL Server Linux node, Log Analytics telemetry, Spot instance support, StopFkh cluster shutdown, Azure Container Registry
Source: Freddys blog, community blog. Summary, key points and quotes are derived (CONTENT-NOTICE.md); read the original for the full text.