Architecture
Lumina’s data flow, reconciliation loops, and caching architecture
Concepts
Understand how Lumina works under the hood
Lumina introduces several key concepts that differentiate it from traditional cost reporting tools: a rate-based cost model that provides instantaneous dollar-per-hour snapshots, multi-account reconciliation that discovers Savings Plans and Reserved Instances across your entire AWS organization, and an in-memory caching architecture that keeps cost data fresh with minimal API overhead.
Architecture
The Architecture page covers Lumina’s data flow from AWS APIs through reconciliation loops into Prometheus metrics. It explains the multi-tier caching system, how Kubernetes nodes are correlated with EC2 instances, and how to configure multi-cluster deployments.
Cost Calculation
The Cost Calculation page details the algorithms Lumina uses to allocate Savings Plans and Reserved Instances to running instances. It covers the strict priority order (Spot, RI, EC2 Instance SP, Compute SP, On-Demand), partial coverage handling, and documents known differences from AWS’s actual billing behavior.
Cost Calculation
How Lumina calculates per-instance costs using Savings Plans and Reserved Instance allocation