Glossary (Cost Optimization)

An operational discipline and cultural practice that brings together engineering, finance and business to make informed economic decisions about cloud resources. FinOps Foundation: "FinOps is a portmanteau of 'Finance' and 'DevOps'. It’s the practice of bringing financial accountability to the variable-spend model of cloud."

Total costs over the full lifecycle of an infrastructure or workload. Includes: infrastructure costs, license costs, FTE operational effort, FTE development effort, vendor management, exit costs and opportunity costs. In the WAF++ context, TCO is the mandatory basis for every cost impact assessment in ADRs.

Long-term economic burdens arising from past architecture decisions that were not fully assessed at the time of the decision. Analogous to technical debt, but measurable in monthly cloud costs. Examples: HA without SLO basis, infinite retention, lock-in without exit plan. Documented in the cost debt register and reviewed quarterly with the architecture board.

A versioned document (YAML or Markdown) in the repository that records all known architectural cost debts. Mandatory fields per entry: ID, description, owner, estimated annual impact (€), status (monitoring/paydown/accepted), paydown plan or acceptance rationale.

Process of clearly attributing cloud costs to workloads, teams, projects and environments. Fundamental prerequisite for all optimization measures. Technically implemented through tagging and cloud provider billing tools (AWS Cost Allocation Tags, Azure Cost Management, GCP Labels + Billing).

Organizational unit to which cloud costs are attributed. Typically a team, a department or a product area. In WAF++ a mandatory tag on all cloud resources.

Adjustment of resource dimensions (CPU, RAM, instance type, storage tier) to actual demand based on measured utilization data (P95/P99). Goal: elimination of over-provisioning without violating SLO requirements.

Upfront commitment for cloud resources (typically 1 or 3 years) in exchange for price discounts of 20–60% compared to on-demand prices. Suitable for baseline workloads with >= 70% utilization. Risk: unused RIs still incur costs.

More flexible alternative to reserved instances (AWS): commitment to a spending amount/hour instead of specific instance types. Enables rightsizing without RI loss.

Low-cost cloud capacity (up to 90% discount) from unused provider pool that can be reclaimed at any time. Suitable for interruptible workloads (batch, CI/CD jobs, stateless services).

Standard pricing model without commitment. Highest price, maximum flexibility. For variable and unpredictable workloads; not for baseline production without RI/SP.

Outbound network traffic from a cloud context to the public internet or to other cloud providers. Typically billed per GB by cloud providers; can become a significant cost position for data-intensive workloads.

Phenomenon where large datasets become "heavy" – workloads tend to develop close to the data, because data transfer is expensive and slow. Important strategic consideration in multi-cloud and migration decisions.

Network connection that enables access to cloud services (S3, KMS, etc.) via the internal cloud network, without an internet gateway. Eliminates internet egress costs for cloud-to-cloud communication within the same provider.

Distributed network of edge servers that caches static content (images, CSS, JS, videos) closer to the end user. Reduces egress costs from the origin and improves latency.

Internal cost allocation process in which cloud costs are actually charged back to the causing teams/projects. Strong incentive for cost ownership.

Like chargeback, but informational only – teams see their costs but are not actually charged. First step toward a cost-awareness culture.

Automation rule for cloud storage resources that moves objects after defined time periods to cheaper storage tiers (e.g. S3 Standard → S3 Infrequent Access → S3 Glacier) or deletes them upon expiry. Essential tool against infinite retention cost debt.

Storage and log tiering concept: Hot (frequent access, highest cost), Warm (infrequent access), Cold (very rare, cheap), Archive (compliance long-term archiving, lowest access need). In WAF++ a mandatory concept for log retention (WAF-COST-070).

Assessment (1–5) used in the WAF++ ADR process for the vendor lock-in risk of an architecture decision. Score >= 4 requires architecture board approval. Score 1 = minimal lock-in (standard APIs, open formats). Score 5 = strong lock-in (proprietary format, no standardized alternatives, high exit costs).

Versioned document that documents an architecture decision with context, options, decision and consequences. In WAF++, every ADR with infrastructure impact contains a mandatory section for the cost impact assessment (WAF-COST-050).

Structured economic assessment of an architecture decision. Mandatory component of the WAF++ ADR process. Contains: TCO estimate, lock-in risk score, data transfer cost estimate, operational effort (FTE), exit cost estimate, 3-year NPV.

Utilization rate of purchased reserved instances. Below 80% means: paid capacity is not being used – classic Architectural Cost Debt from incorrect commitment planning.

Automated cloud provider tools that generate rightsizing recommendations for compute resources based on utilization data. Input for monthly FinOps reviews (WAF-COST-030).

Non-profit organization that standardizes the FinOps framework, certifications and best practices for cloud financial management. Website: https://www.finops.org

Present value calculation of future cash flows. In the WAF++ cost impact assessment, a 3-year NPV is calculated that accounts for the time value component of infrastructure costs. Negative NPVs without corresponding business value are indicators of high Architectural Cost Debt.